Analysis CATegories

Load packages

library(tidyverse)
library(janitor)
library(cowplot)
library(here)
library(readxl)
library(Matrix)
library(lme4)
library(lmerTest)
library(TOSTER)
library(eyetrackingR)
library(wesanderson)
library(gghalves)
library(car)
library(multilevelTools)
library(JWileymisc)

theme_set(theme_cowplot())

knitr::opts_chunk$set(cache = FALSE, warn = FALSE,warning=FALSE, message = FALSE)

Load data

data_file_path <- here::here("data_analysis","registered_report","data","processed_data","CATegories_exp2_processed_data_anonymized.csv")

d <- read_csv(data_file_path)

## add general category properties from typicality dataset
## typicality
animal_rating_names <- read.csv(here::here("data_analysis","registered_report","data","processed_data","animal_ratings_stimuli_full.csv"))
animal_stims <- unique(c(unique(d$left_image),unique(d$right_image)))
typicality <- read.csv(here::here("data_analysis","registered_report","data","processed_data","typicality_animals_summarized.csv")) %>%
  mutate(item_name=str_remove(animal_name,pattern=" ")) %>%
  left_join(animal_rating_names) %>%
  mutate(image_name_resized = str_replace(image_experiment_name,".jpg","")) %>%
  filter(image_name_resized %in% animal_stims)

#read in demographics
demographics <- read.csv(here::here("data_analysis","registered_report","data","processed_data","CATegories_exp2_deidentified_demographics.csv"))

d <- d %>%
  left_join(typicality %>% select(image_name_resized,category,typicality_subjective),by=c("target_image" = "image_name_resized")) %>%
  left_join(demographics)%>%
  rename(typicality_condition=typicality_subjective)%>%
  filter(395<age & age<577) #filter out children outside of the age bounds (older than 13 months and younger than 19 months)

Useable trial summary

In order for a trial to be included, participants must contribute at least 50% looking during the windows of interest when computing baseline-corrected proportion target looking: the critical window (300 ms - 2800 ms relative to target word onset) and the baseline window (-2000 ms - 0 ms relative to target word onset).

critical_window <- c(300,2800)
baseline_window <- c(-2000,0)

summarize_subj_useable_trials_critical_window <- d %>%
  filter(corrected_time_centered>=critical_window[1]&corrected_time_centered<=critical_window[2]) %>%
  group_by(sub_num,age,age_mo, child_gender, session,trial_order,trial_number,target_image,target_typicality_z,condition) %>%
  summarize(
    length_critical_window=n(),
    useable_frames_critical_window=sum(!is.na(accuracy_transformed)),
    percent_useable_critical_window=useable_frames_critical_window/length_critical_window,
    useable_critical_window=ifelse(percent_useable_critical_window>=0.5,1,0), #useable if at least 50% looking
    mean_target_looking_critical=mean(accuracy_transformed,na.rm=TRUE)
  )

summarize_subj_useable_trials_baseline_window <- d %>%
  filter(corrected_time_centered>=baseline_window[1] & corrected_time_centered<=baseline_window[2]) %>%
  group_by(sub_num, session,age,age_mo, child_gender, trial_order,trial_number,target_image,target_typicality_z,condition) %>%
  summarize(
    length_baseline_window=n(),
    useable_frames_baseline_window=sum(!is.na(accuracy_transformed)),
    percent_useable_baseline_window=useable_frames_baseline_window/length_baseline_window,
    useable_baseline_window=ifelse(percent_useable_baseline_window>=0.5,1,0), #useable if at least 50% looking
    mean_target_looking_baseline=mean(accuracy_transformed,na.rm=TRUE)
  )

#overall useable trials
summarize_subj_useable_trials <- summarize_subj_useable_trials_critical_window %>%
  left_join(summarize_subj_useable_trials_baseline_window) %>%
  mutate(
    useable_window = ifelse(useable_baseline_window==1&useable_critical_window==1,1,0),
    corrected_target_looking = mean_target_looking_critical - mean_target_looking_baseline
  )
  

summarize_useable_trials <- summarize_subj_useable_trials %>%
  group_by(sub_num, age, child_gender, session,trial_order) %>%
  summarize(
    num_useable_trials=sum(useable_window),
    num_useable_trials_critical_window = sum(useable_critical_window)
  )

#total trials
summarize_subj_trials <- summarize_useable_trials %>%
  ungroup() %>%
  group_by(sub_num) %>%
  summarize(
    session_num = n(),
    total_trials = sum(num_useable_trials),
    total_trials_critical_window = sum(num_useable_trials_critical_window),
    exclude_participant = ifelse(total_trials<24,1,0),
    exclude_participant_critical = ifelse(total_trials_critical_window<24,1,0)
  )

#average trials contributed
mean(summarize_subj_trials$total_trials)

## [1] 29.11278

#participants to exclude based on data contribution
sum(summarize_subj_trials$exclude_participant)

## [1] 48

#join with main data frame
summarize_useable_trials <- summarize_useable_trials %>%
  left_join(summarize_subj_trials)
d <- d %>%
  left_join(summarize_useable_trials) %>%
  left_join(summarize_subj_useable_trials)

summarize_useable_trials_wide <- summarize_useable_trials %>%
  ungroup() %>%
  select(sub_num,session_num,total_trials,exclude_participant,session,num_useable_trials) %>%
  group_by(sub_num,session_num,total_trials,exclude_participant) %>%
  pivot_wider(
    names_from = "session",
    names_prefix = "num_trials_session_",
    values_from = "num_useable_trials"
  )

#write out useable trial summary
write_csv(summarize_useable_trials_wide,here::here("data_analysis","registered_report","data","processed_data","CATegories_exp2_useable_trial_summary.csv"))

Overall, among the trials contributed by the 133 participants, 82.5% of trials contained sufficient looking to meet our trial-level inclusion criteria (at least 50% looking during both the baseline window and the critical window). 85 of the 133 participants contributed sufficient looking data on at least half of the experimental trials (overall M = 29.1)

Summarize participant information

#summarize subj info
subj_info_multisession <- d %>%
  distinct(sub_num, age,months,age_mo,child_gender,trial_order) %>%
  mutate(
    age_mo_c = age_mo - mean(age_mo),
    age_c = age - mean(age)
  )

subj_info <- d %>%
  distinct(sub_num,child_gender) %>%
  summarize(
    N = n(),
    N_female = sum(child_gender=="f")
  )

overall_subj_info <- subj_info_multisession %>%
  summarize(
    N = length(unique(sub_num)),
    sessions = n(),
    mean_age = mean(age_mo),
    min_age = min(age),
    max_age = max(age),
    sd_age = sd(age_mo)
  ) %>%
  left_join(subj_info)
  
overall_subj_info %>%
  knitr::kable()

N	sessions	mean_age	min_age	max_age	sd_age	N_female
133	234	15.6906	410	574	1.541277	67

# subjects with usable trial data only

subj_info_multisession_usable_trials <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num, age,months,age_mo,child_gender,trial_order) %>%
  mutate(
    age_mo_c = age_mo - mean(age_mo),
    age_c = age - mean(age)
  )

subj_info_usable_trials <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num,child_gender) %>%
  summarize(
    N = n(),
    N_female = sum(child_gender=="f")
  )

overall_subj_info_usable_trials <- subj_info_multisession_usable_trials %>%
  summarize(
    N = length(unique(sub_num)),
    sessions = n(),
    mean_age = mean(age_mo),
    min_age = min(age),
    max_age = max(age),
    sd_age = sd(age_mo)
  ) %>%
  left_join(subj_info_usable_trials)
  
overall_subj_info_usable_trials %>%
  knitr::kable()

N	sessions	mean_age	min_age	max_age	sd_age	N_female
85	166	15.72289	410	574	1.494136	52

= ## Demographics of final usable sample

demographics_summary <- d %>%
  filter(exclude_participant == 0) %>%
  filter(useable_window == 1) %>%
  distinct(sub_num,demographic__us_race_ethnicity_identification,demographic__education_level,demographic__annual_income,demographic__country,demographic__state,demographic__density)


race_ethnicity <- demographics_summary%>%
  group_by(demographic__us_race_ethnicity_identification)%>%
  summarize(
    N = n()
  )

#quick visualiztion
ggplot(demographics_summary,aes(demographic__us_race_ethnicity_identification)) +
  geom_bar()+
  theme(axis.text.x = element_text(angle = 90))

income <- demographics_summary %>%
  group_by(demographic__annual_income)%>%
  summarize(
    N = n()
  )
sum(income$N)

## [1] 86

19/87

## [1] 0.2183908

#quick visualiztion
ggplot(demographics_summary,aes(demographic__annual_income)) +
  geom_bar()+
  theme(axis.text.x = element_text(angle = 90))

education <- demographics_summary %>%
  group_by(demographic__education_level)%>%
  summarize(
    N = n()
  )

states <- demographics_summary %>%
  summarize(
    N = length(unique(demographic__state)),
  )

ggplot(demographics_summary,aes(demographic__state)) +
  geom_bar()+
  theme(axis.text.x = element_text(angle = 90))

population_density <- demographics_summary %>%
  group_by(demographic__density) %>%
  summarize(
    N = n()
  )

Overall

Here, we summarize each participants’ average accuracy during the critical window and average baseline-corrected proportion target looking.

# critical window only
## trial-level
trial_critical_window_accuracy <- d %>%
  filter(exclude_participant_critical==0) %>%
  filter(useable_critical_window==1) %>%
  filter(corrected_time_centered>=300&corrected_time_centered<=2800) %>%
  group_by(sub_num, age,age_mo, child_gender, trial_order,trial_number,category,target_image,target_typicality_z,condition) %>%
  summarize(mean_accuracy=mean(accuracy_transformed,na.rm=TRUE))
## average
avg_critical_window_accuracy <- trial_critical_window_accuracy %>%
  ungroup() %>%
  group_by(sub_num, child_gender) %>%
  summarize(N=n(),
            mean_age = mean(age),
            mean_age_mo = mean(age_mo),
            accuracy=mean(mean_accuracy,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(mean_accuracy,na.rm=T)/sqrt(N),
            lower_ci=accuracy-ci,
            upper_ci=accuracy+ci)

#baseline-corrected target looking
## trial-level
trial_corrected_accuracy <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num,session, age,age_mo, child_gender, trial_order,trial_number,category,target_image,target_typicality_z, condition,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking)

## average
avg_corrected_target_looking <- trial_corrected_accuracy  %>%
  group_by(sub_num, child_gender) %>%
  summarize(N=n(),
            average_corrected_target_looking=mean(corrected_target_looking,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking,na.rm=T)/sqrt(N),
            lower_ci=average_corrected_target_looking-ci,
            upper_ci=average_corrected_target_looking+ci)

By Typicality Condition

Here, we summarize each participants’ average accuracy during the critical window and average baseline-corrected proportion target looking.

# critical window only
avg_critical_window_accuracy_by_typicality <- d %>%
  filter(exclude_participant_critical==0) %>%
  filter(useable_critical_window==1) %>%
  filter(corrected_time_centered>=300&corrected_time_centered<=2800) %>%
  group_by(sub_num, age,age_mo, child_gender, trial_order,trial_number,target_image,target_typicality_z,condition) %>%
  summarize(mean_accuracy=mean(accuracy_transformed,na.rm=TRUE)) %>%
  ungroup() %>%
  group_by(sub_num, child_gender,condition) %>%
  summarize(N=n(),
            mean_age = mean(age),
            mean_age_mo = mean(age_mo),
            accuracy=mean(mean_accuracy,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(mean_accuracy,na.rm=T)/sqrt(N),
            lower_ci=accuracy-ci,
            upper_ci=accuracy+ci)

#baseline-corrected target looking
avg_corrected_target_looking_by_typicality <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num, age,age_mo, child_gender, trial_order,trial_number,target_image,target_typicality_z,condition,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking) %>%
  group_by(sub_num, child_gender,age,age_mo,condition) %>%
  summarize(N=n(),
            average_corrected_target_looking=mean(corrected_target_looking,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking,na.rm=T)/sqrt(N),
            lower_ci=average_corrected_target_looking-ci,
            upper_ci=average_corrected_target_looking+ci)

#avg typicality baseline-corrected target looking
avg_corrected_target_looking_category <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num, months,age_mo,child_gender, trial_order,typicality_condition,trial_number,category,target_image,target_typicality_z,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking) %>%
  group_by(sub_num, age_mo,child_gender, trial_order,typicality_condition,category) %>%
  summarize(N=n(),
            average_corrected_target_looking=mean(corrected_target_looking,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking,na.rm=T)/sqrt(N),
            lower_ci=average_corrected_target_looking-ci,
            upper_ci=average_corrected_target_looking+ci)

#avg typicality baseline-corrected target looking
avg_corrected_target_looking_typicality <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num, months,age_mo,child_gender, trial_order,typicality_condition,trial_number,target_image,target_typicality_z,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking) %>%
  group_by(sub_num, age_mo,child_gender, trial_order,typicality_condition) %>%
  summarize(N=n(),
            average_corrected_target_looking=mean(corrected_target_looking,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking,na.rm=T)/sqrt(N),
            se=sd(corrected_target_looking,na.rm=T)/sqrt(N),
            lower_ci=average_corrected_target_looking-ci,
            upper_ci=average_corrected_target_looking+ci,
            lower_se=average_corrected_target_looking-se,
            upper_se=average_corrected_target_looking+se)

Models

Aim 1

1.1

avg_corrected_target_looking_by_typicality <- avg_corrected_target_looking_by_typicality %>%
  mutate(
    typicality_condition_c = case_when(
      condition == "atypical" ~ -0.5,
      condition == "typical" ~ 0.5,
      TRUE ~ NA_real_
    ),
    typicality_condition_typ = case_when(
      condition == "atypical" ~ -1,
      condition == "typical" ~ 0,
      TRUE ~ NA_real_
    ),
    typicality_condition_atyp = case_when(
      condition == "atypical" ~ 0,
      condition == "typical" ~ 1,
      TRUE ~ NA_real_
    ),
  )

m_1_1 <- lme4::lmer(average_corrected_target_looking ~ 1 + typicality_condition_c + (1|sub_num),data=avg_corrected_target_looking_by_typicality)

summary(m_1_1)

## Linear mixed model fit by REML ['lmerMod']
## Formula: average_corrected_target_looking ~ 1 + typicality_condition_c +  
##     (1 | sub_num)
##    Data: avg_corrected_target_looking_by_typicality
## 
## REML criterion at convergence: -410.7
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.6277 -0.6652 -0.0126  0.6075  3.8645 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.002831 0.05321 
##  Residual             0.014054 0.11855 
## Number of obs: 331, groups:  sub_num, 85
## 
## Fixed effects:
##                        Estimate Std. Error t value
## (Intercept)            0.068604   0.008721   7.867
## typicality_condition_c 0.023856   0.013035   1.830
## 
## Correlation of Fixed Effects:
##             (Intr)
## typclty_cn_ 0.003

confint(m_1_1,method="Wald")

##                               2.5 %     97.5 %
## .sig01                           NA         NA
## .sigma                           NA         NA
## (Intercept)             0.051511065 0.08569668
## typicality_condition_c -0.001690936 0.04940374

1.1.1.

Yes, infants significantly recognized the target word.

1.1.2

No significant effect of typicality

Equivalence test - can’t reject equivalence test

overall_condition_summary <- avg_corrected_target_looking_by_typicality %>%
  group_by(sub_num) %>%
  summarize(
    condition_diff = average_corrected_target_looking[condition=="typical"]-average_corrected_target_looking[condition=="atypical"]
  ) %>%
  ungroup() %>%
  summarize(
    N=n(),
    diff = mean(condition_diff),
    sd = sd(condition_diff)
  )

tsum_TOST(m1=overall_condition_summary$diff,sd1=overall_condition_summary$sd,n1=overall_condition_summary$N,eqb=0.25, eqbound_type = "SMD")

## 
## One-sample t-Test
## 
## The equivalence test was non-significant, t(165) = -1.277, p = 1.02e-01
## The null hypothesis test was non-significant, t(165) = 1.944, p = 5.37e-02
## NHST: don't reject null significance hypothesis that the effect is equal to zero 
## TOST: don't reject null equivalence hypothesis
## 
## TOST Results 
##                 t  df p.value
## t-test      1.944 165   0.054
## TOST Lower  5.165 165 < 0.001
## TOST Upper -1.277 165   0.102
## 
## Effect Sizes 
##            Estimate      SE             C.I. Conf. Level
## Raw         0.02427 0.01249 [0.0036, 0.0449]         0.9
## Hedges's g  0.15016 0.07805 [0.0221, 0.2777]         0.9
## Note: SMD confidence intervals are an approximation. See vignette("SMD_calcs").

1.1.3

m_1_1_3_typ <- lmer(average_corrected_target_looking ~ 1 + typicality_condition_typ + (1|sub_num),data=avg_corrected_target_looking_by_typicality)

summary(m_1_1_3_typ)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: average_corrected_target_looking ~ 1 + typicality_condition_typ +  
##     (1 | sub_num)
##    Data: avg_corrected_target_looking_by_typicality
## 
## REML criterion at convergence: -410.7
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.6277 -0.6652 -0.0126  0.6075  3.8645 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.002831 0.05321 
##  Residual             0.014054 0.11855 
## Number of obs: 331, groups:  sub_num, 85
## 
## Fixed effects:
##                           Estimate Std. Error        df t value Pr(>|t|)    
## (Intercept)                0.08053    0.01090 182.47072   7.388 5.16e-12 ***
## typicality_condition_typ   0.02386    0.01303 244.80317   1.830   0.0684 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## typclty_cn_ 0.600

confint(m_1_1_3_typ,method="Wald")

##                                 2.5 %     97.5 %
## .sig01                             NA         NA
## .sigma                             NA         NA
## (Intercept)               0.059167816 0.10189633
## typicality_condition_typ -0.001690936 0.04940374

Infants successfully recognize words in the typical condition.

m_1_1_3_atyp <- lmer(average_corrected_target_looking ~ 1 + typicality_condition_atyp + (1|sub_num),data=avg_corrected_target_looking_by_typicality)

summary(m_1_1_3_atyp)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: average_corrected_target_looking ~ 1 + typicality_condition_atyp +  
##     (1 | sub_num)
##    Data: avg_corrected_target_looking_by_typicality
## 
## REML criterion at convergence: -410.7
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.6277 -0.6652 -0.0126  0.6075  3.8645 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.002831 0.05321 
##  Residual             0.014054 0.11855 
## Number of obs: 331, groups:  sub_num, 85
## 
## Fixed effects:
##                            Estimate Std. Error        df t value Pr(>|t|)    
## (Intercept)                 0.05668    0.01087 181.46521   5.212 5.05e-07 ***
## typicality_condition_atyp   0.02386    0.01303 244.80317   1.830   0.0684 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## typclty_cn_ -0.597

confint(m_1_1_3_atyp,method="Wald")

##                                  2.5 %     97.5 %
## .sig01                              NA         NA
## .sigma                              NA         NA
## (Intercept)                0.035363003 0.07798834
## typicality_condition_atyp -0.001690936 0.04940374

Infants successfully recognize words in the atypical condition.

1.2

trial_corrected_accuracy <- trial_corrected_accuracy %>%
  mutate(
    typicality_condition_c = case_when(
      condition == "atypical" ~ -0.5,
      condition == "typical" ~ 0.5,
      TRUE ~ NA_real_
    ),
    typicality_condition_typ = case_when(
      condition == "atypical" ~ -1,
      condition == "typical" ~ 0,
      TRUE ~ NA_real_
    ),
    typicality_condition_atyp = case_when(
      condition == "atypical" ~ 0,
      condition == "typical" ~ 1,
      TRUE ~ NA_real_
    ),
  )

m_1_2 <- lmer(corrected_target_looking ~ 1 + typicality_condition_c + 
            (1 + typicality_condition_c||sub_num) +
            (1|category),
          data=trial_corrected_accuracy)
summary(m_1_2)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: corrected_target_looking ~ 1 + typicality_condition_c + (1 +  
##     typicality_condition_c || sub_num) + (1 | category)
##    Data: trial_corrected_accuracy
## 
## REML criterion at convergence: 2294.3
## 
## Scaled residuals: 
##      Min       1Q   Median       3Q      Max 
## -3.14859 -0.63208 -0.02503  0.67538  2.76490 
## 
## Random effects:
##  Groups    Name                   Variance  Std.Dev.
##  sub_num   (Intercept)            0.0025303 0.05030 
##  sub_num.1 typicality_condition_c 0.0000000 0.00000 
##  category  (Intercept)            0.0001651 0.01285 
##  Residual                         0.1191854 0.34523 
## Number of obs: 3138, groups:  sub_num, 85; category, 4
## 
## Fixed effects:
##                         Estimate Std. Error        df t value Pr(>|t|)    
## (Intercept)            7.025e-02  1.047e-02 5.571e+00   6.709 0.000723 ***
## typicality_condition_c 1.663e-02  1.234e-02 3.065e+03   1.348 0.177910    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## typclty_cn_ -0.002
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

confint(m_1_2,method="Wald")

##                               2.5 %    97.5 %
## .sig01                           NA        NA
## .sig02                           NA        NA
## .sig03                           NA        NA
## .sigma                           NA        NA
## (Intercept)             0.049728443 0.0907766
## typicality_condition_c -0.007556538 0.0408094

1.3

Resampling

In order to plot participants’ average proportion looking to the target across the trial, we smooth/ resample time. This is necessary when plotting the timecourses given the variable sampling rate in the data (otherwise the mean observations “jump around” due to varying contributing data composition at different time points).

target_ms_per_frame=1000/30
#adapted from: https://github.com/langcog/peekds/blob/master/R/generate_aoi.R
resample_trial <- function(df_trial) {
  t_origin <- df_trial$corrected_time_centered
  data_origin <- df_trial$accuracy_transformed

  # create the new timestamps for resampling
  t_start <- min(t_origin) - (min(t_origin) %% target_ms_per_frame)
  t_resampled <- seq(from = t_start, to = max(t_origin),
                     by = target_ms_per_frame)

  # exchange strings values with integers for resampling
  # this step critical for interpolating missing vals quickly and correctly
  aoi_num <- data_origin %>%
    dplyr::recode(.missing = 2) #recode NA as 2
    
  # start resampling with approx
  aoi_resampled <- stats::approx(x = t_origin, y = aoi_num, xout = t_resampled,
                                 method = "constant", rule = 2,
                                 ties = "ordered")$y
  aoi_resampled_recoded <- aoi_resampled %>%
    dplyr::recode("0"="0","1"="1","2" = "missing") %>%
    as.numeric()

  # adding back the columns to match schema
  dplyr::tibble(corrected_time_centered = t_resampled,
                accuracy_transformed = aoi_resampled_recoded,
                trial_number = df_trial$trial_number[1],
                sub_num = df_trial$sub_num[1])
}

d_resampled <- d %>%
  dplyr::mutate(sub_num_trial_number = paste(.data$sub_num,
                                           .data$trial_number, sep = "_")) %>%
      split(.$sub_num_trial_number) %>%
      purrr::map_df(resample_trial) %>%
      dplyr::arrange(.data$sub_num, .data$trial_number)

d_info <- d %>%
  select(-corrected_time_centered,-accuracy_transformed) %>%
  distinct(sub_num, exclude_participant, useable_window, age,age_mo, child_gender, trial_order, condition, trial_order,trial_number,target_image,target_typicality_z)

d_resampled <- d_resampled %>%
  left_join(d_info) %>%
  mutate(corrected_time_centered =round(corrected_time_centered,0))

Timecourse analysis - cluster-based permutation analysis

Next, we prepare the data for use with the eyetrackingR package

d_eyetrackingr <- d_resampled %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  unite("unique_trial",trial_order,trial_number,sep="_",remove=FALSE) %>%
  mutate(
    target = case_when(
      is.na(accuracy_transformed) ~ NA,
      accuracy_transformed == 1 ~ TRUE,
      accuracy_transformed == 0 ~ FALSE,
    ),
    distractor = case_when(
      is.na(accuracy_transformed) ~ NA,
      accuracy_transformed == 0 ~ TRUE,
      accuracy_transformed == 1 ~ FALSE,
    ),
    trackloss = case_when(
      is.na(accuracy_transformed) ~ TRUE,
      TRUE ~ FALSE
    )
  ) %>%
  make_eyetrackingr_data(
    participant_column = "sub_num",
    trial_column = "unique_trial",
    time_column = "corrected_time_centered",
    trackloss_column = "trackloss",
    aoi_columns = c("target","distractor"),
    treat_non_aoi_looks_as_missing = TRUE
  )

response_window <- subset_by_window(
  d_eyetrackingr,
  window_start_time = 300,
  window_end_time = 2800,
  rezero=FALSE
)
summary_data_loss <- describe_data(response_window, 'target', 'sub_num')

response_time <- make_time_sequence_data(response_window,
                                  time_bin_size = 100, 
                                  predictor_columns = c("condition"),
                                  aois = "target",
                                  summarize_by = "sub_num" )

# visualize timecourse
plot(response_time, predictor_column = "condition") + 
  theme_light() +
  coord_cartesian(ylim = c(0,1))

#divergence analysis
# tb_analysis <- analyze_time_bins(data = response_time, predictor_column = "condition", test= 'boot_splines', within_subj = TRUE, bs_samples = 1000, alpha = .05/num_time_bins)
# plot(tb_analysis) + theme_light()
# summary(tb_analysis)

#bootstrapped cluster-based permutation analysis
n_samples <- 1000
threshold_t <- 2
df_timeclust <- make_time_cluster_data(response_time, 
                                      test= "t.test", paired=TRUE,
                                      predictor_column = "condition", 
                                      threshold = threshold_t) 
plot(df_timeclust) +  ylab("T-Statistic") + theme_light()

summary(df_timeclust)

## Test Type:    t.test 
## Predictor:    condition 
## Formula:  Pair(Prop[condition == "atypical"], Prop[condition == "typical"]) ~      1 
## Summary of Clusters ======
## [1] Cluster   Direction EndTime  
## <0 rows> (or 0-length row.names)

clust_analysis <- analyze_time_clusters(df_timeclust, within_subj=TRUE, paired=TRUE,
                                        samples=n_samples)
plot(clust_analysis) + theme_light()

summary(clust_analysis)

## Test Type:    t.test 
## Predictor:    condition 
## Formula:  Pair(Prop[condition == "atypical"], Prop[condition == "typical"]) ~      1 
## Null Distribution   ====== 
##  Mean:        0.1028 
##  2.5%:        -11.7908 
## 97.5%:        11.4885 
## Summary of Clusters ======
## [1] Cluster     Direction   EndTime     Probability
## <0 rows> (or 0-length row.names)

Timecourse Plot

Next, we plot the data. First we summarize the data in two steps: (1) summarize the data by subject for each time point, followed by (2) averaging looking for each time point across subjects.

#summarizing within subject for each time point
summarize_subj <- d_resampled %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  group_by(sub_num, age,age_mo, child_gender, trial_order, corrected_time_centered) %>%
  summarize(N=n(),
            non_na_n = sum(!is.na(accuracy_transformed)), 
            mean_accuracy=mean(accuracy_transformed,na.rm=TRUE),
            ci=qt(0.975, non_na_n-1)*sd(accuracy_transformed,na.rm=T)/sqrt(non_na_n),
            lower_ci=mean_accuracy-ci,
            upper_ci=mean_accuracy+ci) %>%
  ungroup()

#summarizing across subjects for each time point
summarize_across_subj <- summarize_subj %>%
  group_by(corrected_time_centered) %>%
  dplyr::summarize(n=n(),
            accuracy=mean(mean_accuracy,na.rm=TRUE),
            sd_accuracy=sd(mean_accuracy,na.rm=TRUE),
            se_accuracy=sd_accuracy/sqrt(n))

ggplot(summarize_across_subj,aes(corrected_time_centered,accuracy))+
  xlim(-2000,4000)+
  geom_smooth(method="gam")+
  geom_errorbar(aes(ymin=accuracy-se_accuracy,ymax=accuracy+se_accuracy),width=0)+
  geom_point()+
  geom_vline(xintercept=0,size=1.5)+
  geom_hline(yintercept=0.5,size=1.2,linetype="dashed")+
  geom_vline(xintercept=300,linetype="dotted")+
  ylim(0.35,0.65)

ggsave(here::here("figures","overall_accuracy.png"))

Timecourse by age

summarize_across_subj_by_age <- summarize_subj %>%
  mutate(age_group=ifelse(age_mo>16,"older than 16 months","younger than 16 months")) %>%
  group_by(age_group,corrected_time_centered) %>%
  dplyr::summarize(n=n(),
                   accuracy=mean(mean_accuracy,na.rm=TRUE),
                   sd_accuracy=sd(mean_accuracy,na.rm=TRUE),
                   se_accuracy=sd_accuracy/sqrt(n))
ggplot(summarize_across_subj_by_age,aes(corrected_time_centered,accuracy))+
  xlim(-2000,4000)+
  geom_smooth(method="gam")+
  geom_errorbar(aes(ymin=accuracy-se_accuracy,ymax=accuracy+se_accuracy),width=0)+
  geom_point()+
  geom_vline(xintercept=0,size=1.5)+
  geom_hline(yintercept=0.5,size=1.2,linetype="dashed")+
  geom_vline(xintercept=300,linetype="dotted")+
  facet_wrap(~age_group)

ggsave(here::here("figures","overall_accuracy_by_age.png"),width=12, height=9)

Timecourse by condition

summarize_subj_condition <- d_resampled %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  group_by(sub_num, age,age_mo, child_gender, condition, trial_order, corrected_time_centered) %>%
  summarize(mean_accuracy=mean(accuracy_transformed,na.rm=TRUE))

summarize_across_subj_cond <- summarize_subj_condition %>%
  group_by(condition,corrected_time_centered) %>%
  summarize(n=n(),
            accuracy=mean(mean_accuracy,na.rm=TRUE),
            sd_accuracy=sd(mean_accuracy,na.rm=TRUE),
            se_accuracy=sd_accuracy/sqrt(n))

num_subjects <- summarize_across_subj_cond %>%
  group_by()%>%
  summarize(max_subnum=max(n))

summarize_across_subj_cond<- summarize_subj_condition %>%
  group_by(condition,corrected_time_centered) %>%
  summarize(n=n(),
            accuracy=mean(mean_accuracy,na.rm=TRUE),
            sd_accuracy=sd(mean_accuracy,na.rm=TRUE),
            se_accuracy=sd_accuracy/sqrt(n))

ggplot(summarize_across_subj_cond,aes(corrected_time_centered,accuracy,color=condition))+
  xlim(-2500,4000)+
  geom_rect(data = data.frame(xmin = 300,
                              xmax = 2800,
                              ymin = -Inf,
                              ymax = Inf),
            aes(x=NULL, y=NULL,xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax,color=NULL),
            fill = "grey", alpha = 0.2)+
  geom_rect(data = data.frame(xmin = -2000,
                              xmax = 0,
                              ymin = -Inf,
                              ymax = Inf),
            aes(x=NULL, y=NULL,xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax,color=NULL),
            fill = "grey", alpha = 0.2)+
  geom_errorbar(aes(ymin=accuracy-se_accuracy,ymax=accuracy+se_accuracy),width=0)+
  geom_point(alpha=0.5)+
  geom_smooth(data=summarize_subj_condition,aes(y=mean_accuracy),method="gam")+
  geom_vline(xintercept=0,size=1.5)+
  geom_hline(yintercept=0.5,size=1.2,linetype="dashed")+
  geom_vline(xintercept=300,linetype="dotted")+
  geom_vline(xintercept=2800,linetype="dotted")+
  geom_vline(xintercept=-2000,linetype="dotted")+
  geom_vline(xintercept=0,linetype="dotted")+
  theme(legend.position = c(0.75,0.15))+
  annotate("text",label="Critical Window",x=1550,y=0.9)+
  annotate("text",label="Baseline Window",x=-1000,y=0.9)+
  ylim(0,1)+
  ylab("Proportion Target Looking")+
  xlab("Time (centered on target word onset, in ms)")

ggsave(here::here("figures","typicality_accuracy.png"),width=10,height=6)

Aim 2

Change across age

trial_corrected_accuracy <- trial_corrected_accuracy %>%
  left_join(subj_info_multisession)
  

m_2 <- lmer(corrected_target_looking ~ 1 + typicality_condition_c * age_mo_c + 
            (1 + typicality_condition_c|sub_num) +
            (1|category),
          data=trial_corrected_accuracy)

summary(m_2)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: corrected_target_looking ~ 1 + typicality_condition_c * age_mo_c +  
##     (1 + typicality_condition_c | sub_num) + (1 | category)
##    Data: trial_corrected_accuracy
## 
## REML criterion at convergence: 2300.4
## 
## Scaled residuals: 
##      Min       1Q   Median       3Q      Max 
## -3.12875 -0.63679 -0.01852  0.67408  2.79689 
## 
## Random effects:
##  Groups   Name                   Variance  Std.Dev. Corr
##  sub_num  (Intercept)            0.0020489 0.04527      
##           typicality_condition_c 0.0002154 0.01468  1.00
##  category (Intercept)            0.0001748 0.01322      
##  Residual                        0.1191112 0.34512      
## Number of obs: 3138, groups:  sub_num, 85; category, 4
## 
## Fixed effects:
##                                  Estimate Std. Error        df t value Pr(>|t|)
## (Intercept)                     6.901e-02  1.032e-02 4.993e+00   6.684  0.00114
## typicality_condition_c          1.574e-02  1.247e-02 9.188e+02   1.262  0.20727
## age_mo_c                        1.614e-02  5.317e-03 8.928e+01   3.036  0.00314
## typicality_condition_c:age_mo_c 6.903e-03  8.418e-03 9.621e+02   0.820  0.41243
##                                   
## (Intercept)                     **
## typicality_condition_c            
## age_mo_c                        **
## typicality_condition_c:age_mo_c   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) typc__ ag_m_c
## typclty_cn_  0.060              
## age_mo_c    -0.049  0.000       
## typclt__:__  0.001 -0.075  0.076
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

confint(m_2,method="Wald")

##                                        2.5 %     97.5 %
## .sig01                                    NA         NA
## .sig02                                    NA         NA
## .sig03                                    NA         NA
## .sig04                                    NA         NA
## .sigma                                    NA         NA
## (Intercept)                      0.048773471 0.08924206
## typicality_condition_c          -0.008706112 0.04018854
## age_mo_c                         0.005720197 0.02656220
## typicality_condition_c:age_mo_c -0.009596693 0.02340258

Quick Visualization

ggplot(avg_corrected_target_looking_by_typicality,aes(age,average_corrected_target_looking,color=condition))+
  geom_point()+
  geom_smooth(method="lm")+
  geom_hline(yintercept=0,linetype="dashed")

ggplot(trial_critical_window_accuracy,aes(age,mean_accuracy,color=condition))+
  geom_point(alpha=0.1)+
  geom_smooth()

m <- lmer(mean_target_looking_critical ~ 1 + typicality_condition_c * age_mo_c + mean_target_looking_baseline +
            (1 + typicality_condition_c|sub_num) +
            (1|category),
          data=trial_corrected_accuracy)
summary(m)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: 
## mean_target_looking_critical ~ 1 + typicality_condition_c * age_mo_c +  
##     mean_target_looking_baseline + (1 + typicality_condition_c |  
##     sub_num) + (1 | category)
##    Data: trial_corrected_accuracy
## 
## REML criterion at convergence: 922.8
## 
## Scaled residuals: 
##      Min       1Q   Median       3Q      Max 
## -2.44570 -0.63233  0.05242  0.68487  1.98760 
## 
## Random effects:
##  Groups   Name                   Variance  Std.Dev. Corr
##  sub_num  (Intercept)            1.760e-03 0.04196      
##           typicality_condition_c 6.906e-05 0.00831  1.00
##  category (Intercept)            9.717e-04 0.03117      
##  Residual                        7.627e-02 0.27617      
## Number of obs: 3138, groups:  sub_num, 85; category, 4
## 
## Fixed effects:
##                                  Estimate Std. Error        df t value Pr(>|t|)
## (Intercept)                     5.145e-01  2.006e-02 6.768e+00  25.646 5.39e-08
## typicality_condition_c          1.632e-02  9.942e-03 1.480e+03   1.642  0.10079
## age_mo_c                        1.517e-02  4.521e-03 8.740e+01   3.356  0.00117
## mean_target_looking_baseline    9.091e-02  2.177e-02 3.094e+03   4.176 3.05e-05
## typicality_condition_c:age_mo_c 5.864e-03  6.710e-03 1.532e+03   0.874  0.38232
##                                    
## (Intercept)                     ***
## typicality_condition_c             
## age_mo_c                        ** 
## mean_target_looking_baseline    ***
## typicality_condition_c:age_mo_c    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) typc__ ag_m_c mn_t__
## typclty_cn_  0.021                     
## age_mo_c    -0.024  0.001              
## mn_trgt_lk_ -0.532 -0.001  0.005       
## typclt__:__ -0.001 -0.075  0.058  0.004
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

Accuracy across age

Critical Window Only

ggplot(filter(avg_critical_window_accuracy,mean_age<700),aes(mean_age,accuracy))+
  geom_pointrange(aes(ymin=lower_ci,ymax=upper_ci), 
                  position=position_jitter(width=0.1),
                  width=0,
                  size=1.5) +
  geom_hline(yintercept=0.5,linetype="dashed")+
  geom_smooth(method="lm")+
  xlab("Age (in days)")+
  ylab("Proportion Target Looking\nduring the Critical Window")+
  ylim(0,1)

ggsave(here::here("figures","age_relationship_critical_window_accuracy.png"),width=7,height=6)

Baseline-corrected proportion target looking

# ggplot(avg_corrected_target_looking,aes(age,average_corrected_target_looking))+
#   geom_pointrange(aes(ymin=lower_ci,ymax=upper_ci), 
#                   position=position_jitter(width=0.1),
#                   width=0,
#                   size=1.5) +
#   geom_hline(yintercept=0,linetype="dashed")+
#   geom_smooth(method="lm")+
#   xlab("Age (in months)")+
#   ylab("Baseline-Corrected Proportion Target Looking")+
#   ylim(-0.55,0.55)+
#   scale_x_continuous(breaks=seq(12,18,1))
# ggsave(here::here("figures","age_relationship_baseline_corrected_accuracy.png"),width=7,height=6)

Target Looking By Item

Next, we investigate item-level (target word) variation in proportion target looking.

Overall target looking during critical window and baseline window

First, we inspect overall target looking in the critical window and in the baseline window. Note the baseline effects, such that dog and cat are more likely to be fixated during baseline than bird and fish.

#average target looking during the baseline and critical window by item for each subject
avg_subj_target_looking_by_item <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num,  age,age_mo, child_gender, trial_order,trial_number,category,target_image,target_typicality_z,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking) %>%
  group_by(sub_num, age,age_mo, child_gender, trial_order,category) %>%
  summarize(N=n(),
            mean_critical_accuracy=mean(mean_target_looking_critical,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(mean_target_looking_critical,na.rm=T)/sqrt(N),
            lower_ci=mean_critical_accuracy-ci,
            upper_ci=mean_critical_accuracy+ci,
            mean_baseline_accuracy=mean(mean_target_looking_baseline,na.rm=TRUE),
            baseline_ci=qt(0.975, N-1)*sd(mean_target_looking_baseline,na.rm=T)/sqrt(N),
            lower_baseline_ci=mean_baseline_accuracy-baseline_ci,
            upper_baseline_ci=mean_baseline_accuracy+baseline_ci)

#summarize average target looking across subjects
avg_target_looking_by_item <- avg_subj_target_looking_by_item %>%
  group_by(category) %>%
  summarize(N=n(),
            critical_accuracy=mean(mean_critical_accuracy,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(mean_critical_accuracy,na.rm=T)/sqrt(N),
            lower_ci=critical_accuracy-ci,
            upper_ci=critical_accuracy+ci,
            baseline_accuracy=mean(mean_baseline_accuracy,na.rm=TRUE),
            baseline_ci=qt(0.975, N-1)*sd(mean_baseline_accuracy,na.rm=T)/sqrt(N),
            lower_baseline_ci=baseline_accuracy-baseline_ci,
            upper_baseline_ci=baseline_accuracy+baseline_ci)
avg_target_looking_by_item %>%
  knitr::kable()

category	N	critical_accuracy	ci	lower_ci	upper_ci	baseline_accuracy	baseline_ci	lower_baseline_ci	upper_baseline_ci
bird	165	0.5345443	0.0231118	0.5114325	0.5576561	0.4563859	0.0174362	0.4389497	0.4738222
cat	166	0.5974269	0.0226906	0.5747363	0.6201174	0.5169224	0.0175366	0.4993858	0.5344590
dog	165	0.5785324	0.0210893	0.5574431	0.5996218	0.5323489	0.0153733	0.5169757	0.5477222
fish	165	0.5204611	0.0246016	0.4958595	0.5450627	0.4620394	0.0168223	0.4452171	0.4788618

Target looking during the baseline and critical window split by age

#summarize average corrected target looking across subject
avg_target_looking_by_item_by_age <- avg_subj_target_looking_by_item %>%
  mutate(age_group=ifelse(age_mo>16,"older than 16 months","younger than 16 months")) %>%
  group_by(age_group,category) %>%
  summarize(N=n(),
            critical_accuracy=mean(mean_critical_accuracy,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(mean_critical_accuracy,na.rm=T)/sqrt(N),
            lower_ci=critical_accuracy-ci,
            upper_ci=critical_accuracy+ci,
            baseline_accuracy=mean(mean_baseline_accuracy,na.rm=TRUE),
            baseline_ci=qt(0.975, N-1)*sd(mean_baseline_accuracy,na.rm=T)/sqrt(N),
            lower_baseline_ci=baseline_accuracy-baseline_ci,
            upper_baseline_ci=baseline_accuracy+baseline_ci)
avg_target_looking_by_item_by_age %>%
  knitr::kable()

age_group	category	N	critical_accuracy	ci	lower_ci	upper_ci	baseline_accuracy	baseline_ci	lower_baseline_ci	upper_baseline_ci
older than 16 months	bird	68	0.5632700	0.0338378	0.5294322	0.5971078	0.4499680	0.0277305	0.4222375	0.4776985
older than 16 months	cat	68	0.5987425	0.0323744	0.5663681	0.6311169	0.4914161	0.0253063	0.4661098	0.5167224
older than 16 months	dog	68	0.5846130	0.0330569	0.5515561	0.6176700	0.5359675	0.0205925	0.5153750	0.5565601
older than 16 months	fish	68	0.5720050	0.0333565	0.5386485	0.6053615	0.4666964	0.0304112	0.4362852	0.4971076
younger than 16 months	bird	97	0.5144067	0.0312134	0.4831933	0.5456200	0.4608851	0.0227758	0.4381094	0.4836609
younger than 16 months	cat	98	0.5965140	0.0316579	0.5648560	0.6281719	0.5346206	0.0236818	0.5109387	0.5583024
younger than 16 months	dog	97	0.5742698	0.0278494	0.5464203	0.6021192	0.5298122	0.0220962	0.5077160	0.5519083
younger than 16 months	fish	97	0.4843272	0.0333021	0.4510251	0.5176294	0.4587747	0.0195579	0.4392169	0.4783326

Overall baseline-corrected proportion target looking

Next, we investigate item-level variation in word recognition as measured by baseline-corrected proportion target looking (to help account for the baseline difference noted above).

#average corrected target looking by item for each subject
avg_subj_corrected_target_looking_by_item <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  mutate(age_group=ifelse(age_mo>16,"older than 16 months","younger than 16 months")) %>%
  distinct(sub_num, months,age_mo,age_group, child_gender, trial_order,trial_number,category,target_image,target_typicality_z,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking) %>%
  group_by(sub_num, age_mo,age_group, child_gender, trial_order,category) %>%
  summarize(N=n(),
            average_corrected_target_looking=mean(corrected_target_looking,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking,na.rm=T)/sqrt(N),
            lower_ci=average_corrected_target_looking-ci,
            upper_ci=average_corrected_target_looking+ci)

#summarize average corrected target looking across subject
avg_corrected_target_looking_by_item <- avg_subj_corrected_target_looking_by_item %>%
  group_by(category) %>%
  summarize(N=n(),
            corrected_target_looking=mean(average_corrected_target_looking,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(average_corrected_target_looking,na.rm=T)/sqrt(N),
            lower_ci=corrected_target_looking-ci,
            upper_ci=corrected_target_looking+ci)
avg_corrected_target_looking_by_item %>%
  knitr::kable()

category	N	corrected_target_looking	ci	lower_ci	upper_ci
bird	165	0.0781583	0.0287984	0.0493599	0.1069568
cat	166	0.0805045	0.0279757	0.0525288	0.1084802
dog	165	0.0461835	0.0255777	0.0206059	0.0717612
fish	165	0.0584217	0.0279178	0.0305038	0.0863395

Baseline-corrected proportion target looking split by age

#summarize average corrected target looking across subject
avg_corrected_target_looking_by_item_by_age <- avg_subj_corrected_target_looking_by_item %>%
  group_by(age_group,category) %>%
  summarize(N=n(),
            corrected_target_looking=mean(average_corrected_target_looking,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(average_corrected_target_looking,na.rm=T)/sqrt(N),
            lower_ci=corrected_target_looking-ci,
            upper_ci=corrected_target_looking+ci)
avg_corrected_target_looking_by_item_by_age %>%
  knitr::kable()

age_group	category	N	corrected_target_looking	ci	lower_ci	upper_ci
older than 16 months	bird	68	0.1133020	0.0449011	0.0684009	0.1582031
older than 16 months	cat	68	0.1073264	0.0421662	0.0651602	0.1494926
older than 16 months	dog	68	0.0486455	0.0375555	0.0110900	0.0862010
older than 16 months	fish	68	0.1053086	0.0424541	0.0628545	0.1477627
younger than 16 months	bird	97	0.0535215	0.0374115	0.0161100	0.0909331
younger than 16 months	cat	98	0.0618934	0.0374395	0.0244539	0.0993328
younger than 16 months	dog	97	0.0444576	0.0351927	0.0092649	0.0796503
younger than 16 months	fish	97	0.0255525	0.0361833	-0.0106308	0.0617358

Baseline-corrected proportion target looking split by typicality

#average corrected target looking by item by typicality for each subject
avg_subj_corrected_target_looking_by_item_by_typicality <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num, months,age_mo, trial_order,trial_number,typicality_condition,category,target_image,target_typicality_z,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking) %>%
  group_by(sub_num, age_mo, trial_order,typicality_condition,category) %>%
  summarize(N=n(),
            average_corrected_target_looking=mean(corrected_target_looking,na.rm=TRUE))
#summarize average corrected target looking across subject
avg_corrected_target_looking_by_item_by_typicality <- avg_subj_corrected_target_looking_by_item_by_typicality %>%
  group_by(typicality_condition,category) %>%
  summarize(N=n(),
            corrected_target_looking=mean(average_corrected_target_looking,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(average_corrected_target_looking,na.rm=T)/sqrt(N),
            lower_ci=corrected_target_looking-ci,
            upper_ci=corrected_target_looking+ci)
avg_corrected_target_looking_by_item_by_typicality %>%
  knitr::kable()

typicality_condition	category	N	corrected_target_looking	ci	lower_ci	upper_ci
atypical	bird	159	0.0983027	0.0397753	0.0585274	0.1380780
atypical	cat	163	0.0586166	0.0413179	0.0172987	0.0999345
atypical	dog	162	0.0165194	0.0378078	-0.0212884	0.0543272
atypical	fish	164	0.0601608	0.0368548	0.0233060	0.0970156
typical	bird	163	0.0657351	0.0363182	0.0294169	0.1020533
typical	cat	163	0.1129400	0.0396868	0.0732532	0.1526269
typical	dog	164	0.0818268	0.0374507	0.0443761	0.1192775
typical	fish	159	0.0585323	0.0392131	0.0193191	0.0977454

Baseline-corrected proportion target looking by category

pal <- wes_palette("Rushmore1", n=5)
set.seed(1)
jitterer <- position_jitter(width = .05,seed=1)
p2 <- ggplot(avg_corrected_target_looking_category,aes(x=typicality_condition,y=average_corrected_target_looking, fill=typicality_condition))+
  geom_half_violin(data=filter(avg_corrected_target_looking_category, typicality_condition=="atypical"),position = position_nudge(x = -.1, y = 0), width=1,trim = FALSE, alpha = .8,color=NA,side="l")+
  geom_half_violin(data=filter(avg_corrected_target_looking_category, typicality_condition=="typical"),position = position_nudge(x = .1, y = 0), width=1,trim = FALSE, alpha = .8,color=NA,side="r")+
  geom_path(aes(group=sub_num),color="black",fill=NA,alpha=0.05,size=0.75,position=jitterer)+
  geom_point(aes(color=typicality_condition,group=sub_num), size = 2.5, alpha=0.05,position=jitterer)+
  geom_point(data=avg_corrected_target_looking_by_item_by_typicality,aes(y=corrected_target_looking),color="black",size=1.8)+
  geom_line(data=avg_corrected_target_looking_by_item_by_typicality,aes(y=corrected_target_looking,group=1),color="black",size=1.5)+
  geom_errorbar(data=avg_corrected_target_looking_by_item_by_typicality,aes(y=corrected_target_looking,ymin=lower_ci,ymax=upper_ci),width=0,color="black")+
  #geom_boxplot(outlier.shape = NA, alpha = .5, width = .1, colour = "black")+
  #scale_colour_brewer(palette = "Dark2")+
  #scale_fill_brewer(palette = "Dark2")+
  geom_hline(yintercept=0,linetype="dashed")+
  scale_colour_manual(values=pal[c(3,4)])+
  scale_fill_manual(values=pal[c(3,4)])+
  facet_wrap(.~category)+
  theme(legend.position="none")+
  xlab("Typicality Condition")+
  ylab("Baseline-Corrected\nProportion Target Looking")+
  theme(axis.title.x = element_text(face="bold", size=20),
        axis.text.x  = element_text(size=16),
        axis.title.y = element_text(face="bold", size=20),
        axis.text.y  = element_text(size=16),
        strip.text.x = element_text(size = 16,face="bold"))


ggsave(here::here("figures","age_relationship_baseline_corrected_accuracy_category.png"),width=7,height=6)

pal <- wes_palette("Rushmore1", n=5)
p1 <- ggplot(avg_corrected_target_looking_typicality,aes(age_mo,average_corrected_target_looking,color=typicality_condition,group=typicality_condition))+
  geom_pointrange(aes(ymin=lower_se,ymax=upper_se), 
                  position=position_jitter(width=0.2),
                  width=0,
                  size=1) +
  geom_hline(yintercept=0,linetype="dashed")+
  geom_smooth(method="lm",color="black",size=1.3)+
  xlab("Age (in months)")+
  ylab("Baseline-Corrected\nProportion Target Looking")+
  #ylim(-0.55,0.5)+
  scale_colour_manual(values=pal[c(3,4)])+
  scale_x_continuous(breaks=seq(12,18,1))+
  facet_wrap(~typicality_condition)+
  theme(
    strip.background = element_rect(size=1, colour = "black"),
    strip.text = element_text(size=16,face="bold"),
    axis.title=element_text(size=20,face="bold"),
    axis.text = element_text(size=14))+
  theme(legend.position="none")
ggsave(here::here("figures","age_relationship_baseline_corrected_accuracy_typicality.png"),width=9,height=6)

plot_grid(p1,p2,nrow=1,ncol=2,labels=c("A","B"),size=24)

ggsave(here::here("figures","age_category_typicality.png"),width=12,height=6)

Aim 3

To test whether individual differences in word recognition or typicality effects are predicted by differences in experiences with each exemplar.

3.1 z-scored by participant

#zscore parent report of typicality within participants
parent_typicality_z <- d %>%
  group_by(sub_num) %>%
  mutate(target_parent_typicality_z = ((target_parent_typicality_rating - mean(target_parent_typicality_rating))/sd(target_parent_typicality_rating)),
         distractor_parent_typicality_z = ((distractor_parent_typicality_rating - mean(distractor_parent_typicality_rating))/sd(distractor_parent_typicality_rating)))
  
parent_typicality <- parent_typicality_z %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num,age_mo,age,child_gender,condition, trial_order,trial_number,category,target_image,target_typicality_z,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking,target_parent_typicality_z,distractor_parent_typicality_z,target_image,target_parent_typicality_rating,distractor_parent_typicality_rating) %>%
  mutate(
    age_mo_c = age_mo - mean(age_mo),
    age_c = age - mean(age)
  )%>%
  group_by(sub_num,target_image, age_mo_c,age, age_c, age_mo, condition, category,target_parent_typicality_z,target_parent_typicality_rating,distractor_parent_typicality_z) %>%
  summarize(N=n(),
            average_corrected_target_looking=mean(corrected_target_looking,na.rm=TRUE)) %>%
  na.omit(target_parent_typicality_rating)

#subject details for aim 3 analysis (how many participants have survey data)
aim3_subject_info <- parent_typicality %>%
  ungroup()%>%
  summarize(
    N = length(unique(sub_num)),
    mean_age = mean(age_c),
    sd_age = sd(age_c)
  )

aim3_subject_info%>%
  knitr::kable()

N	mean_age	sd_age
72	0	4.177312

#model
m_3_1 <- lmer(average_corrected_target_looking ~ 1 + target_parent_typicality_z + age_mo_c + (target_parent_typicality_z||sub_num) + (1|category), parent_typicality)

summary(m_3_1)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: average_corrected_target_looking ~ 1 + target_parent_typicality_z +  
##     age_mo_c + (target_parent_typicality_z || sub_num) + (1 |      category)
##    Data: parent_typicality
## 
## REML criterion at convergence: 1962.9
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.1768 -0.6325 -0.0276  0.6730  2.7533 
## 
## Random effects:
##  Groups    Name                       Variance Std.Dev.
##  sub_num   (Intercept)                0.002377 0.04876 
##  sub_num.1 target_parent_typicality_z 0.000000 0.00000 
##  category  (Intercept)                0.000263 0.01622 
##  Residual                             0.118664 0.34448 
## Number of obs: 2694, groups:  sub_num, 72; category, 4
## 
## Fixed effects:
##                              Estimate Std. Error         df t value Pr(>|t|)   
## (Intercept)                 7.187e-02  1.198e-02  4.994e+00   6.001  0.00185 **
## target_parent_typicality_z  7.778e-03  6.706e-03  2.187e+03   1.160  0.24620   
## age_mo_c                   -2.643e-02  4.854e-02  2.618e+03  -0.545  0.58611   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) trg___
## trgt_prnt__  0.000       
## age_mo_c     0.000 -0.007
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

confint(m_3_1,method="Wald")

##                                   2.5 %     97.5 %
## .sig01                               NA         NA
## .sig02                               NA         NA
## .sig03                               NA         NA
## .sigma                               NA         NA
## (Intercept)                 0.048400194 0.09534746
## target_parent_typicality_z -0.005364784 0.02092160
## age_mo_c                   -0.121578099 0.06870914

Exploratory analyses to see why age isn’t a predictor in model 3

#cor.test(parent_typicality$target_parent_typicality_z~parent_typicality$age_c)

ggplot(parent_typicality, aes(target_parent_typicality_z,age)) +
  geom_point()+
  geom_smooth(method = "lm")

ggplot(d, aes(condition,child_age_rounded))+
  geom_point()+
  stat_summary(fun = "mean")

parent_typicality <- parent_typicality %>%
  mutate(
    typicality_condition_c = case_when(
      condition == "atypical" ~ -0.5,
      condition == "typical" ~ 0.5,
      TRUE ~ NA_real_
    ),
    typicality_condition_typ = case_when(
      condition == "atypical" ~ -1,
      condition == "typical" ~ 0,
      TRUE ~ NA_real_
    ),
    typicality_condition_atyp = case_when(
      condition == "atypical" ~ 0,
      condition == "typical" ~ 1,
      TRUE ~ NA_real_
    ),
  )

ggplot(parent_typicality, aes(condition,target_parent_typicality_rating))+
  geom_point()#+facet_wrap(~category)

  #geom_bar(stat = "identity")


ggplot(parent_typicality, aes(target_parent_typicality_z))+
  geom_histogram()+ facet_wrap(~condition)#+

  #facet_wrap(~parent_typicality$category)

Check normality assumptions

qqPlot(parent_typicality$target_parent_typicality_z)

## [1]  504 1454

Check distribution of participants between Aim 2 & Aim 3

aim2_compare_sub_nums <- trial_corrected_accuracy %>%
  distinct(sub_num,age,age_mo)

range(aim2_compare_sub_nums$age_mo, na.rm = T)

## [1] 13.5 18.9

aim3_compare_sub_nums <- parent_typicality %>%
  ungroup()%>%
  distinct(sub_num,age,age_mo)

diff <- anti_join(aim2_compare_sub_nums, aim3_compare_sub_nums, by = c("sub_num" = "sub_num"))

Z-scored by category by participant experience ratings

#zscore parent report of typicality within participants by 
parent_typicality_category_z <- d %>%
  group_by(sub_num,category) %>%
  mutate(target_parent_typicality_category_z = ((target_parent_typicality_rating - mean(target_parent_typicality_rating))/sd(target_parent_typicality_rating)),
         distractor_parent_typicality_category_z = ((distractor_parent_typicality_rating - mean(distractor_parent_typicality_rating))/sd(distractor_parent_typicality_rating)))
  
parent_typicality_category <- parent_typicality_category_z %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num,months,age,child_gender, trial_order,trial_number,category,target_image,target_parent_typicality_category_z,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking,distractor_parent_typicality_category_z,target_image) %>%
  mutate(age_c = age - mean(age,na.rm = T))%>%
  group_by(sub_num,target_image, category,target_parent_typicality_category_z, age_c) %>%
  summarize(N=n(),
            average_corrected_target_looking=mean(corrected_target_looking,na.rm=TRUE)) %>%
  na.omit(target_parent_typicality_rating)

#model
m_3_1_2 <- lmer(average_corrected_target_looking ~ target_parent_typicality_category_z+age_c+ (target_parent_typicality_category_z||sub_num) + (1|category), parent_typicality_category)

summary(m_3_1_2)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: 
## average_corrected_target_looking ~ target_parent_typicality_category_z +  
##     age_c + (target_parent_typicality_category_z || sub_num) +  
##     (1 | category)
##    Data: parent_typicality_category
## 
## REML criterion at convergence: 1987.2
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.1600 -0.6210 -0.0277  0.6770  2.7540 
## 
## Random effects:
##  Groups    Name                                Variance  Std.Dev.
##  sub_num   (Intercept)                         0.0023069 0.04803 
##  sub_num.1 target_parent_typicality_category_z 0.0000000 0.00000 
##  category  (Intercept)                         0.0003026 0.01739 
##  Residual                                      0.1191765 0.34522 
## Number of obs: 2703, groups:  sub_num, 73; category, 4
## 
## Fixed effects:
##                                       Estimate Std. Error         df t value
## (Intercept)                          7.183e-02  1.233e-02  4.715e+00   5.825
## target_parent_typicality_category_z  2.694e-03  6.643e-03  2.637e+03   0.406
## age_c                               -4.698e-04  1.603e-03  2.626e+03  -0.293
##                                     Pr(>|t|)   
## (Intercept)                          0.00256 **
## target_parent_typicality_category_z  0.68511   
## age_c                                0.76940   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) tr____
## trgt_prn___ -0.002       
## age_c        0.000 -0.004
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

confint(m_3_1_2,method="Wald")

##                                            2.5 %      97.5 %
## .sig01                                        NA          NA
## .sig02                                        NA          NA
## .sig03                                        NA          NA
## .sigma                                        NA          NA
## (Intercept)                          0.047664623 0.096003603
## target_parent_typicality_category_z -0.010326531 0.015714763
## age_c                               -0.003610816 0.002671119

Model accounting for distractor rating too

The outcome variable, average_corrected_target_looking accounts for baseline preferences for different images. We have typicality information for both the targets and the distractor. Does accounting for the saliency/experience with the distractor predict accuracy?

#model
m_3_3 <- lmer(average_corrected_target_looking ~ target_parent_typicality_z+distractor_parent_typicality_z+age_c+ (target_parent_typicality_z||sub_num) + (1|category), parent_typicality)

summary(m_3_3)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: average_corrected_target_looking ~ target_parent_typicality_z +  
##     distractor_parent_typicality_z + age_c + (target_parent_typicality_z ||  
##     sub_num) + (1 | category)
##    Data: parent_typicality
## 
## REML criterion at convergence: 1976.9
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.1867 -0.6340 -0.0237  0.6720  2.7570 
## 
## Random effects:
##  Groups    Name                       Variance Std.Dev.
##  sub_num   (Intercept)                0.002381 0.04879 
##  sub_num.1 target_parent_typicality_z 0.000000 0.00000 
##  category  (Intercept)                0.000254 0.01594 
##  Residual                             0.118675 0.34449 
## Number of obs: 2694, groups:  sub_num, 72; category, 4
## 
## Fixed effects:
##                                  Estimate Std. Error         df t value
## (Intercept)                     7.187e-02  1.188e-02  5.035e+00   6.048
## target_parent_typicality_z      4.513e-03  7.550e-03  1.645e+03   0.598
## distractor_parent_typicality_z  6.985e-03  7.434e-03  2.297e+03   0.940
## age_c                          -7.013e-04  1.589e-03  2.617e+03  -0.441
##                                Pr(>|t|)   
## (Intercept)                     0.00174 **
## target_parent_typicality_z      0.55014   
## distractor_parent_typicality_z  0.34750   
## age_c                           0.65907   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) trg___ dst___
## trgt_prnt__  0.000              
## dstrctr_p__  0.000 -0.460       
## age_c        0.000  0.000 -0.012
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

confint(m_3_3,method="Wald")

##                                       2.5 %      97.5 %
## .sig01                                   NA          NA
## .sig02                                   NA          NA
## .sig03                                   NA          NA
## .sigma                                   NA          NA
## (Intercept)                     0.048582657 0.095166078
## target_parent_typicality_z     -0.010285755 0.019311033
## distractor_parent_typicality_z -0.007585312 0.021556198
## age_c                          -0.003816418 0.002413795

No, accounting for experience with the distractor does not explain infants’ accuracy.

Quick Visualization

#by subject
ggplot(parent_typicality,aes(target_parent_typicality_z,average_corrected_target_looking))+
  geom_point(alpha=0.1)+
  geom_smooth(method = "lm")+
  facet_wrap(~condition)

#by category
ggplot(parent_typicality_category,aes(target_parent_typicality_category_z,average_corrected_target_looking))+
  geom_point(alpha=0.1)+
  geom_smooth(method = "lm")+
  facet_wrap(~category)

Parent’s ratings of their infant’s experience with each exemplar does not predict accuracy.

Aim 4

We will conduct a series of analyses to determine whether our results hold across a variety of different analytic decisions

4.1. Window

Define alternative critical window

critical_window_alternative <- c(300,1800)

summarize_subj_useable_trials_critical_window_alternative <- d %>%
  filter(corrected_time_centered>=critical_window_alternative[1]&corrected_time_centered<=critical_window_alternative[2]) %>%
  group_by(sub_num,age,age_mo, child_gender, session,trial_order,trial_number,target_image,target_typicality_z,condition) %>%
  summarize(
    length_critical_window_alternative=n(),
    useable_frames_critical_window_alternative=sum(!is.na(accuracy_transformed)),
    percent_useable_critical_window_alternative=useable_frames_critical_window_alternative/length_critical_window_alternative,
    useable_critical_window_alternative=ifelse(percent_useable_critical_window_alternative>=0.5,1,0), #useable if at least 50% looking
    mean_target_looking_critical_alternative=mean(accuracy_transformed,na.rm=TRUE)
  )

#overall useable trials
summarize_subj_useable_trials_alternative <- summarize_subj_useable_trials_critical_window_alternative %>%
  left_join(summarize_subj_useable_trials_baseline_window) %>%
  mutate(
    useable_window_alternative = ifelse(useable_baseline_window==1&useable_critical_window_alternative==1,1,0),
    corrected_target_looking_alternative = mean_target_looking_critical_alternative - mean_target_looking_baseline
  )
  

summarize_useable_trials_alternative <- summarize_subj_useable_trials_alternative %>%
  group_by(sub_num, age, child_gender, session,trial_order) %>%
  summarize(
    num_useable_trials_alternative=sum(useable_window_alternative),
    num_useable_trials_critical_window_alternative = sum(useable_critical_window_alternative)
  )

#total trials
summarize_subj_trials_alternative <- summarize_useable_trials_alternative %>%
  ungroup() %>%
  group_by(sub_num) %>%
  summarize(
    session_num_alternative = n(),
    total_trials_alternative = sum(num_useable_trials_alternative),
    total_trials_critical_window_alternative = sum(num_useable_trials_critical_window_alternative),
    exclude_participant_alternative = ifelse(total_trials_alternative<24,1,0),
    exclude_participant_critical_alternative = ifelse(total_trials_critical_window_alternative<24,1,0)
  )

#average trials contributed
mean(summarize_subj_trials_alternative$total_trials_alternative)

## [1] 29.52632

#participants to exclude based on data contribution
sum(summarize_subj_trials_alternative$exclude_participant_alternative)

## [1] 49

#join with main data frame
summarize_useable_trials_alternative <- summarize_useable_trials_alternative %>%
  left_join(summarize_subj_trials_alternative)
d <- d %>%
  left_join(summarize_useable_trials_alternative) %>%
  left_join(summarize_subj_useable_trials_alternative)

summarize_useable_trials_wide_alternative <- summarize_useable_trials_alternative %>%
  ungroup() %>%
  select(sub_num,session_num_alternative,total_trials_alternative,exclude_participant_alternative,session,num_useable_trials_alternative) %>%
  group_by(sub_num,session_num_alternative,total_trials_alternative,exclude_participant_alternative) %>%
  pivot_wider(
    names_from = "session",
    names_prefix = "num_trials_session_alternative",
    values_from = "num_useable_trials_alternative"
  )

By Typicality Condition for alternative critical window

Here, we summarize each participants’ average accuracy during the alternative critical window and average baseline-corrected proportion target looking.

# critical window only
avg_critical_window_accuracy_by_typicality_alternative <- d %>%
  filter(exclude_participant_critical_alternative==0) %>%
  filter(useable_critical_window_alternative==1) %>%
  filter(corrected_time_centered>=300&corrected_time_centered<=1800) %>%
  group_by(sub_num, age,age_mo, child_gender, trial_order,trial_number,target_image,target_typicality_z,condition) %>%
  summarize(mean_accuracy_alternative=mean(accuracy_transformed,na.rm=TRUE)) %>%
  ungroup() %>%
  group_by(sub_num, child_gender,condition) %>%
  summarize(N=n(),
            mean_age = mean(age),
            mean_age_mo = mean(age_mo),
            accuracy=mean(mean_accuracy_alternative,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(mean_accuracy_alternative,na.rm=T)/sqrt(N),
            lower_ci=accuracy-ci,
            upper_ci=accuracy+ci)

#baseline-corrected target looking
avg_corrected_target_looking_by_typicality_alternative <- d %>%
  filter(exclude_participant_alternative==0) %>%
  filter(useable_window_alternative==1) %>%
  distinct(sub_num, age,age_mo, child_gender, trial_order,trial_number,target_image,target_typicality_z,condition,mean_target_looking_critical_alternative,mean_target_looking_baseline,corrected_target_looking_alternative) %>%
  group_by(sub_num, child_gender,age,age_mo,condition) %>%
  summarize(N=n(),
            average_corrected_target_looking_alternative=mean(corrected_target_looking_alternative,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking_alternative,na.rm=T)/sqrt(N),
            lower_ci_alternative=average_corrected_target_looking_alternative-ci,
            upper_ci_alternative=average_corrected_target_looking_alternative+ci)

#avg typicality baseline-corrected target looking
avg_corrected_target_looking_category_alternative <- d %>%
  filter(exclude_participant_alternative==0) %>%
  filter(useable_window_alternative==1) %>%
  distinct(sub_num, months,age_mo,child_gender, trial_order,typicality_condition,trial_number,category,target_image,target_typicality_z,mean_target_looking_critical_alternative,mean_target_looking_baseline,corrected_target_looking_alternative) %>%
  group_by(sub_num, age_mo,child_gender, trial_order,typicality_condition,category) %>%
  summarize(N=n(),
            average_corrected_target_looking_alternative=mean(corrected_target_looking_alternative,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking_alternative,na.rm=T)/sqrt(N),
            lower_ci_alternative=average_corrected_target_looking_alternative-ci,
            upper_ci_alternative=average_corrected_target_looking_alternative+ci)

#avg typicality baseline-corrected target looking
avg_corrected_target_looking_typicality_alternative <- d %>%
  filter(exclude_participant_alternative==0) %>%
  filter(useable_window_alternative==1) %>%
  distinct(sub_num, months,age_mo,child_gender, trial_order,typicality_condition,trial_number,target_image,target_typicality_z,mean_target_looking_critical_alternative,mean_target_looking_baseline,corrected_target_looking_alternative) %>%
  group_by(sub_num, age_mo,child_gender, trial_order,typicality_condition) %>%
  summarize(N=n(),
            average_corrected_target_looking_alternative=mean(corrected_target_looking_alternative,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking_alternative,na.rm=T)/sqrt(N),
            se=sd(corrected_target_looking_alternative,na.rm=T)/sqrt(N),
            lower_ci_alternative=average_corrected_target_looking_alternative-ci,
            upper_ci_alternative=average_corrected_target_looking_alternative+ci,
            lower_se_alternative=average_corrected_target_looking_alternative-se,
            upper_se_alternative=average_corrected_target_looking_alternative+se)

#baseline-corrected target looking
## trial-level
trial_corrected_accuracy_alternative <- d %>%
  filter(exclude_participant_alternative==0) %>%
  filter(useable_window_alternative==1) %>%
  distinct(sub_num,session, age,age_mo, child_gender, trial_order,trial_number,category,target_image,target_typicality_z, condition,mean_target_looking_critical_alternative,mean_target_looking_baseline,corrected_target_looking_alternative)
## average
avg_corrected_target_looking_alternative <- trial_corrected_accuracy_alternative  %>%
  group_by(sub_num, child_gender) %>%
  summarize(N=n(),
            average_corrected_target_looking_alternative=mean(corrected_target_looking_alternative,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking_alternative,na.rm=T)/sqrt(N),
            lower_ci=average_corrected_target_looking_alternative-ci,
            upper_ci=average_corrected_target_looking_alternative+ci)

4.1.1.1

Does typicality influence target looking?

avg_corrected_target_looking_by_typicality_alternative <- avg_corrected_target_looking_by_typicality_alternative %>%
  mutate(
    typicality_condition_c = case_when(
      condition == "atypical" ~ -0.5,
      condition == "typical" ~ 0.5,
      TRUE ~ NA_real_
    ),
    typicality_condition_typ = case_when(
      condition == "atypical" ~ -1,
      condition == "typical" ~ 0,
      TRUE ~ NA_real_
    ),
    typicality_condition_atyp = case_when(
      condition == "atypical" ~ 0,
      condition == "typical" ~ 1,
      TRUE ~ NA_real_
    ),
  )

m_4_1_1_1 <- lmer(average_corrected_target_looking_alternative ~ 1 + typicality_condition_c + (1|sub_num),data=avg_corrected_target_looking_by_typicality_alternative)

summary(m_4_1_1_1)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: 
## average_corrected_target_looking_alternative ~ 1 + typicality_condition_c +  
##     (1 | sub_num)
##    Data: avg_corrected_target_looking_by_typicality_alternative
## 
## REML criterion at convergence: -365.3
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -4.2752 -0.6062  0.0011  0.6181  2.8029 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.002026 0.04501 
##  Residual             0.016709 0.12926 
## Number of obs: 327, groups:  sub_num, 84
## 
## Fixed effects:
##                         Estimate Std. Error        df t value Pr(>|t|)    
## (Intercept)            6.359e-02  8.687e-03 8.367e+01   7.320 1.39e-10 ***
## typicality_condition_c 2.601e-02  1.430e-02 2.443e+02   1.819   0.0702 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## typclty_cn_ 0.003

confint(m_4_1_1_1,method="Wald")

##                               2.5 %     97.5 %
## .sig01                           NA         NA
## .sigma                           NA         NA
## (Intercept)             0.046561476 0.08061255
## typicality_condition_c -0.002019258 0.05403043

4.1.1.1.2

No significant effect of typicality

Equivalence test - can’t reject equivalence test

overall_condition_summary_alternative <- avg_corrected_target_looking_by_typicality_alternative %>%
  group_by(sub_num) %>%
  summarize(
    condition_diff_alternative = average_corrected_target_looking_alternative[condition=="typical"]-average_corrected_target_looking_alternative[condition=="atypical"]
  ) %>%
  ungroup() %>%
  summarize(
    N=n(),
    diff = mean(condition_diff_alternative),
    sd = sd(condition_diff_alternative)
  )

tsum_TOST(m1=overall_condition_summary_alternative$diff,sd1=overall_condition_summary_alternative$sd,n1=overall_condition_summary_alternative$N,eqb=0.25, eqbound_type = "SMD")

## 
## One-sample t-Test
## 
## The equivalence test was non-significant, t(163) = -1.254, p = 1.06e-01
## The null hypothesis test was non-significant, t(163) = 1.947, p = 5.32e-02
## NHST: don't reject null significance hypothesis that the effect is equal to zero 
## TOST: don't reject null equivalence hypothesis
## 
## TOST Results 
##                 t  df p.value
## t-test      1.947 163   0.053
## TOST Lower  5.149 163 < 0.001
## TOST Upper -1.254 163   0.106
## 
## Effect Sizes 
##            Estimate      SE             C.I. Conf. Level
## Raw         0.02634 0.01353  [0.004, 0.0487]         0.9
## Hedges's g  0.15136 0.07853 [0.0225, 0.2797]         0.9
## Note: SMD confidence intervals are an approximation. See vignette("SMD_calcs").

4.1.1.1.3

m_1_1_3_typ_alternative <- lmer(average_corrected_target_looking_alternative ~ 1 + typicality_condition_typ + (1|sub_num),data=avg_corrected_target_looking_by_typicality_alternative)

summary(m_1_1_3_typ_alternative)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: 
## average_corrected_target_looking_alternative ~ 1 + typicality_condition_typ +  
##     (1 | sub_num)
##    Data: avg_corrected_target_looking_by_typicality_alternative
## 
## REML criterion at convergence: -365.3
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -4.2752 -0.6062  0.0011  0.6181  2.8029 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.002026 0.04501 
##  Residual             0.016709 0.12926 
## Number of obs: 327, groups:  sub_num, 84
## 
## Fixed effects:
##                           Estimate Std. Error        df t value Pr(>|t|)    
## (Intercept)                0.07659    0.01127 202.49093   6.799 1.16e-10 ***
## typicality_condition_typ   0.02601    0.01430 244.27862   1.819   0.0702 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## typclty_cn_ 0.637

confint(m_1_1_3_typ_alternative,method="Wald")

##                                 2.5 %     97.5 %
## .sig01                             NA         NA
## .sigma                             NA         NA
## (Intercept)               0.054510013 0.09866960
## typicality_condition_typ -0.002019258 0.05403043

Infants successfully recognize words in the typical condition.

m_1_1_3_atyp_alternative <- lmer(average_corrected_target_looking_alternative ~ 1 + typicality_condition_atyp + (1|sub_num),data=avg_corrected_target_looking_by_typicality_alternative)

summary(m_1_1_3_atyp_alternative)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: 
## average_corrected_target_looking_alternative ~ 1 + typicality_condition_atyp +  
##     (1 | sub_num)
##    Data: avg_corrected_target_looking_by_typicality_alternative
## 
## REML criterion at convergence: -365.3
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -4.2752 -0.6062  0.0011  0.6181  2.8029 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.002026 0.04501 
##  Residual             0.016709 0.12926 
## Number of obs: 327, groups:  sub_num, 84
## 
## Fixed effects:
##                            Estimate Std. Error        df t value Pr(>|t|)    
## (Intercept)                 0.05058    0.01124 201.48091   4.502 1.14e-05 ***
## typicality_condition_atyp   0.02601    0.01430 244.27862   1.819   0.0702 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## typclty_cn_ -0.634

confint(m_1_1_3_atyp_alternative,method="Wald")

##                                  2.5 %     97.5 %
## .sig01                              NA         NA
## .sigma                              NA         NA
## (Intercept)                0.028563399 0.07260504
## typicality_condition_atyp -0.002019258 0.05403043

Infants successfully recognize words in the atypical condition.

4.1.1.3

Resampling

In order to plot participants’ average proportion looking to the target across the trial, we smooth/ resample time. This is necessary when plotting the timecourses given the variable sampling rate in the data (otherwise the mean observations “jump around” due to varying contributing data composition at different time points).

target_ms_per_frame=1000/30
#adapted from: https://github.com/langcog/peekds/blob/master/R/generate_aoi.R
resample_trial_alt <- function(df_trial_alt) {
  t_origin_alt <- df_trial_alt$corrected_time_centered
  data_origin_alt <- df_trial_alt$accuracy_transformed

  # create the new timestamps for resampling
  t_start_alt <- min(t_origin_alt) - (min(t_origin_alt) %% target_ms_per_frame)
  t_resampled_alt <- seq(from = t_start_alt, to = max(t_origin_alt),
                     by = target_ms_per_frame)

  # exchange strings values with integers for resampling
  # this step critical for interpolating missing vals quickly and correctly
  aoi_num_alt <- data_origin_alt %>%
    dplyr::recode(.missing = 2) #recode NA as 2
    
  # start resampling with approx
  aoi_resampled_alt <- stats::approx(x = t_origin_alt, y = aoi_num_alt, xout = t_resampled_alt,
                                 method = "constant", rule = 2,
                                 ties = "ordered")$y
  aoi_resampled_recoded_alt <- aoi_resampled_alt %>%
    dplyr::recode("0"="0","1"="1","2" = "missing") %>%
    as.numeric()

  # adding back the columns to match schema
  dplyr::tibble(corrected_time_centered = t_resampled_alt,
                accuracy_transformed = aoi_resampled_recoded_alt,
                trial_number = df_trial_alt$trial_number[1],
                sub_num = df_trial_alt$sub_num[1])
}

d_resampled_alt <- d %>%
  dplyr::mutate(sub_num_trial_number = paste(.data$sub_num,
                                           .data$trial_number, sep = "_")) %>%
      split(.$sub_num_trial_number) %>%
      purrr::map_df(resample_trial_alt) %>%
      dplyr::arrange(.data$sub_num, .data$trial_number)

d_info_alt <- d %>%
  select(-corrected_time_centered,-accuracy_transformed) %>%
  distinct(sub_num, exclude_participant_alternative, useable_window_alternative, age,age_mo, child_gender, trial_order, condition, trial_order,trial_number,target_image,target_typicality_z)

d_resampled_alt <- d_resampled_alt %>%
  left_join(d_info_alt) %>%
  mutate(corrected_time_centered =round(corrected_time_centered,0))

Timecourse analysis - cluster-based permutation analysis

Next, we prepare the data for use with the eyetrackingR package

d_eyetrackingr_alt <- d_resampled_alt %>%
  filter(exclude_participant_alternative==0) %>%
  filter(useable_window_alternative==1) %>%
  unite("unique_trial",trial_order,trial_number,sep="_",remove=FALSE) %>%
  mutate(
    target = case_when(
      is.na(accuracy_transformed) ~ NA,
      accuracy_transformed == 1 ~ TRUE,
      accuracy_transformed == 0 ~ FALSE,
    ),
    distractor = case_when(
      is.na(accuracy_transformed) ~ NA,
      accuracy_transformed == 0 ~ TRUE,
      accuracy_transformed == 1 ~ FALSE,
    ),
    trackloss = case_when(
      is.na(accuracy_transformed) ~ TRUE,
      TRUE ~ FALSE
    )
  ) %>%
  make_eyetrackingr_data(
    participant_column = "sub_num",
    trial_column = "unique_trial",
    time_column = "corrected_time_centered",
    trackloss_column = "trackloss",
    aoi_columns = c("target","distractor"),
    treat_non_aoi_looks_as_missing = TRUE
  )

response_window_alt <- subset_by_window(
  d_eyetrackingr_alt,
  window_start_time = 300,
  window_end_time = 1800,
  rezero=FALSE
)
summary_data_loss_alt <- describe_data(response_window_alt, 'target', 'sub_num')

response_time_alt <- make_time_sequence_data(response_window_alt,
                                  time_bin_size = 100, 
                                  predictor_columns = c("condition"),
                                  aois = "target",
                                  summarize_by = "sub_num" )

# visualize timecourse
plot(response_time_alt, predictor_column = "condition") + 
  theme_light() +
  coord_cartesian(ylim = c(0,1))

#divergence analysis
# tb_analysis <- analyze_time_bins(data = response_time, predictor_column = "condition", test= 'boot_splines', within_subj = TRUE, bs_samples = 1000, alpha = .05/num_time_bins)
# plot(tb_analysis) + theme_light()
# summary(tb_analysis)

#bootstrapped cluster-based permutation analysis
n_samples_alt <- 100
threshold_t_alt <- 2
df_timeclust_alt <- make_time_cluster_data(response_time_alt, 
                                      test= "t.test", paired=TRUE,
                                      predictor_column = "condition", 
                                      threshold = threshold_t) 
plot(df_timeclust_alt) +  ylab("T-Statistic") + theme_light()

summary(df_timeclust_alt)

## Test Type:    t.test 
## Predictor:    condition 
## Formula:  Pair(Prop[condition == "atypical"], Prop[condition == "typical"]) ~      1 
## Summary of Clusters ======
## [1] Cluster   Direction EndTime  
## <0 rows> (or 0-length row.names)

clust_analysis_alt <- analyze_time_clusters(df_timeclust_alt, within_subj=TRUE, paired=TRUE,
                                        samples=n_samples)
plot(clust_analysis_alt) + theme_light()

summary(clust_analysis_alt)

## Test Type:    t.test 
## Predictor:    condition 
## Formula:  Pair(Prop[condition == "atypical"], Prop[condition == "typical"]) ~      1 
## Null Distribution   ====== 
##  Mean:        -0.016 
##  2.5%:        -9.2828 
## 97.5%:        10.3631 
## Summary of Clusters ======
## [1] Cluster     Direction   EndTime     Probability
## <0 rows> (or 0-length row.names)

4.1.1.2

Trial level typicality effects

trial_corrected_accuracy_alternative <- trial_corrected_accuracy_alternative %>%
  mutate(
    typicality_condition_c = case_when(
      condition == "atypical" ~ -0.5,
      condition == "typical" ~ 0.5,
      TRUE ~ NA_real_
    ),
    typicality_condition_typ = case_when(
      condition == "atypical" ~ -1,
      condition == "typical" ~ 0,
      TRUE ~ NA_real_
    ),
    typicality_condition_atyp = case_when(
      condition == "atypical" ~ 0,
      condition == "typical" ~ 1,
      TRUE ~ NA_real_
    ),
  )

m_4_1_1_2 <- lmer(corrected_target_looking_alternative ~ 1 + typicality_condition_c + 
            (1 + typicality_condition_c||sub_num) +
            (1|category),
          data=trial_corrected_accuracy_alternative)
summary(m_4_1_1_2)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: corrected_target_looking_alternative ~ 1 + typicality_condition_c +  
##     (1 + typicality_condition_c || sub_num) + (1 | category)
##    Data: trial_corrected_accuracy_alternative
## 
## REML criterion at convergence: 3254.1
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -2.6947 -0.7085  0.0039  0.7366  2.3965 
## 
## Random effects:
##  Groups    Name                   Variance Std.Dev.
##  sub_num   (Intercept)            0.001410 0.03755 
##  sub_num.1 typicality_condition_c 0.000000 0.00000 
##  category  (Intercept)            0.000195 0.01396 
##  Residual                         0.163018 0.40375 
## Number of obs: 3141, groups:  sub_num, 84; category, 4
## 
## Fixed effects:
##                         Estimate Std. Error        df t value Pr(>|t|)   
## (Intercept)            6.450e-02  1.086e-02 4.023e+00   5.941  0.00395 **
## typicality_condition_c 2.007e-02  1.442e-02 3.073e+03   1.392  0.16398   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## typclty_cn_ 0.000 
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

confint(m_4_1_1_2,method="Wald")

##                               2.5 %     97.5 %
## .sig01                           NA         NA
## .sig02                           NA         NA
## .sig03                           NA         NA
## .sigma                           NA         NA
## (Intercept)             0.043224552 0.08578423
## typicality_condition_c -0.008185369 0.04832312

4.1.2

Aim 2 with alternative time window

Change across age

trial_corrected_accuracy_alternative <- trial_corrected_accuracy_alternative %>%
  mutate(
    typicality_condition_c = case_when(
      condition == "atypical" ~ -0.5,
      condition == "typical" ~ 0.5,
      TRUE ~ NA_real_
    ),
    typicality_condition_typ = case_when(
      condition == "atypical" ~ -1,
      condition == "typical" ~ 0,
      TRUE ~ NA_real_
    ),
    typicality_condition_atyp = case_when(
      condition == "atypical" ~ 0,
      condition == "typical" ~ 1,
      TRUE ~ NA_real_
    ),
  )
trial_corrected_accuracy_alternative <- trial_corrected_accuracy_alternative %>%
  left_join(subj_info_multisession)
  

m_4_1_2 <- lmer(corrected_target_looking_alternative ~ 1 + typicality_condition_c * age_mo_c + 
            (1 + typicality_condition_c|sub_num) +
            (1|category),
          data=trial_corrected_accuracy_alternative)
summary(m_4_1_2)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: corrected_target_looking_alternative ~ 1 + typicality_condition_c *  
##     age_mo_c + (1 + typicality_condition_c | sub_num) + (1 |      category)
##    Data: trial_corrected_accuracy_alternative
## 
## REML criterion at convergence: 3262.6
## 
## Scaled residuals: 
##      Min       1Q   Median       3Q      Max 
## -2.69543 -0.70411 -0.00237  0.74166  2.43239 
## 
## Random effects:
##  Groups   Name                   Variance  Std.Dev. Corr
##  sub_num  (Intercept)            1.024e-03 0.032004     
##           typicality_condition_c 4.124e-05 0.006422 1.00
##  category (Intercept)            2.045e-04 0.014299     
##  Residual                        1.630e-01 0.403752     
## Number of obs: 3141, groups:  sub_num, 84; category, 4
## 
## Fixed effects:
##                                  Estimate Std. Error        df t value Pr(>|t|)
## (Intercept)                     6.289e-02  1.077e-02 3.735e+00   5.839  0.00529
## typicality_condition_c          1.924e-02  1.449e-02 2.104e+03   1.328  0.18428
## age_mo_c                        1.452e-02  5.432e-03 8.884e+01   2.673  0.00895
## typicality_condition_c:age_mo_c 6.834e-03  9.796e-03 2.139e+03   0.698  0.48551
##                                   
## (Intercept)                     **
## typicality_condition_c            
## age_mo_c                        **
## typicality_condition_c:age_mo_c   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) typc__ ag_m_c
## typclty_cn_  0.016              
## age_mo_c    -0.063  0.003       
## typclt__:__  0.003 -0.089  0.019
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

confint(m_4_1_2,method="Wald")

##                                        2.5 %     97.5 %
## .sig01                                    NA         NA
## .sig02                                    NA         NA
## .sig03                                    NA         NA
## .sig04                                    NA         NA
## .sigma                                    NA         NA
## (Intercept)                      0.041781958 0.08400266
## typicality_condition_c          -0.009155022 0.04764379
## age_mo_c                         0.003873203 0.02516793
## typicality_condition_c:age_mo_c -0.012366573 0.02603415

Age is still a significant predictor when we use a smaller critical window. There is no age by typicality condition interaction.

4.1.3

Aim 3: test whether individual differences in word recognition or typicality effects are predicted by differences in experiences with each exemplar using the alternative critical window

3.1 z-scored by participant

#zscore parent report of typicality within participants
parent_typicality_z_alt <- d %>%
  group_by(sub_num) %>%
  mutate(target_parent_typicality_z = ((target_parent_typicality_rating - mean(target_parent_typicality_rating))/sd(target_parent_typicality_rating)),
         distractor_parent_typicality_z = ((distractor_parent_typicality_rating - mean(distractor_parent_typicality_rating))/sd(distractor_parent_typicality_rating)))
  
parent_typicality_alt <- parent_typicality_z_alt %>%
  filter(exclude_participant_alternative==0) %>%
  filter(useable_window_alternative==1) %>%
  distinct(sub_num,months,age,child_gender, trial_order,trial_number,category,target_image,target_typicality_z,mean_target_looking_critical_alternative,mean_target_looking_baseline,corrected_target_looking_alternative,target_parent_typicality_z,distractor_parent_typicality_z,target_image,target_parent_typicality_rating,distractor_parent_typicality_rating) %>%
  mutate(age_c = age - mean(age,na.rm = T))%>%
  group_by(sub_num,target_image, age_c, category,target_parent_typicality_z,target_parent_typicality_rating,distractor_parent_typicality_z) %>%
  summarize(N=n(),
            average_corrected_target_looking_alt=mean(corrected_target_looking_alternative,na.rm=TRUE)) %>%
  na.omit(target_parent_typicality_rating)

#subject details for aim 3 analysis (how many participants have survey data)
aim3_subject_info_alt <- parent_typicality_alt %>%
  ungroup()%>%
  summarize(
    N = length(unique(sub_num)),
    mean_age = mean(age_c),
    sd_age = sd(age_c)
  )

aim3_subject_info_alt%>%
  knitr::kable()

N	mean_age	sd_age
71	0	4.194607

#model
m_4_1_3_1 <- lmer(average_corrected_target_looking_alt ~ target_parent_typicality_z+age_c+ (target_parent_typicality_z||sub_num) + (1|category), parent_typicality_alt)

summary(m_4_1_3_1)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: average_corrected_target_looking_alt ~ target_parent_typicality_z +  
##     age_c + (target_parent_typicality_z || sub_num) + (1 | category)
##    Data: parent_typicality_alt
## 
## REML criterion at convergence: 2775.2
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -2.7339 -0.6871  0.0137  0.7333  2.4426 
## 
## Random effects:
##  Groups    Name                       Variance  Std.Dev.
##  sub_num   (Intercept)                0.0011487 0.03389 
##  sub_num.1 target_parent_typicality_z 0.0000000 0.00000 
##  category  (Intercept)                0.0004104 0.02026 
##  Residual                             0.1618511 0.40231 
## Number of obs: 2687, groups:  sub_num, 71; category, 4
## 
## Fixed effects:
##                             Estimate Std. Error        df t value Pr(>|t|)   
## (Intercept)                6.869e-02  1.339e-02 3.584e+00   5.128  0.00916 **
## target_parent_typicality_z 1.385e-02  7.851e-03 2.220e+03   1.764  0.07780 . 
## age_c                      4.088e-04  1.851e-03 2.614e+03   0.221  0.82519   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) trg___
## trgt_prnt__  0.000       
## age_c        0.000 -0.006
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

confint(m_4_1_3_1,method="Wald")

##                                   2.5 %      97.5 %
## .sig01                               NA          NA
## .sig02                               NA          NA
## .sig03                               NA          NA
## .sigma                               NA          NA
## (Intercept)                 0.042439501 0.094943635
## target_parent_typicality_z -0.001535489 0.029240075
## age_c                      -0.003218484 0.004036101

Parent ratings are not significantly related to baseline corrected target looking with the alternative critical window.

4.2 excluding unknown words

We registered using CDI responses as a way to remove unknown words; however we did not administer the CDI. Thus, we are not including this analysis since it would require an arbitrary cutoff for word recognition as a proxy for understanding a word (i.e., what is the difference between 50% accuracy and 50.01% accuracy)

4.3 Reaction time as the dependent measure

We will fit models analogous to those in 1.1 and 1.2 using reaction time as our primary dependent measure rather than accuracy

Load RT (computed using compute_RT.R & rt_helper.R)

rt_path <- here::here("data_analysis","registered_report","data","processed_data","CATegories_exp2_RT_by_trial.csv")
d_rt <- read_csv(rt_path)

d_trial_level <- d %>%
  distinct(sub_num,session, trial_number,condition, age, age_mo, target_image, child_gender, category)

d_rt<- d_trial_level %>%
  left_join(d_rt)

#d_rt<- trial_corrected_accuracy %>%
 # left_join(d_rt)

#participants must contribute 4 typical and 4 atypical trials to be included in analysis
d_rt_subj_summary <- d_rt %>%
  #filter(shift_type == "D-T")%>%
  group_by(sub_num, condition,shift_type) %>%
  summarize(
    trials = n(),
    useable_trials=ifelse(trials>=4,1,0)
  )

#add exclusionary criteria to DF
d_rt <- d_rt %>%
  left_join(d_rt_subj_summary)

Visualization of RTs

hist(filter(d_rt, shift_type=="D-T")$shift_start_rt)

The data are right skewed, which is common for RTs. We will use log transformations in the subsequent models to account for the distribution of the data.

Models

4.3.1 Subject-level RT

avg_subj_RT <- d_rt %>%
  filter(shift_type=="D-T")%>%
  filter(useable_trials=="1")%>%
  group_by(sub_num, child_gender,condition) %>%
  summarize(N=n(),
            average_RT=mean(rt,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(rt,na.rm=T)/sqrt(N),
            lower_ci=average_RT-ci,
            upper_ci=average_RT+ci)


avg_subj_RT <- avg_subj_RT %>%
  mutate(
    typicality_condition_c = case_when(
      condition == "atypical" ~ -0.5,
      condition == "typical" ~ 0.5,
      TRUE ~ NA_real_
    ),
    typicality_condition_typ = case_when(
      condition == "atypical" ~ -1,
      condition == "typical" ~ 0,
      TRUE ~ NA_real_
    ),
    typicality_condition_atyp = case_when(
      condition == "atypical" ~ 0,
      condition == "typical" ~ 1,
      TRUE ~ NA_real_
    ),
  )

m_4_1 <- lmer(log(average_RT) ~ 1 + typicality_condition_c + (1|sub_num),data=avg_subj_RT)

summary(m_4_1)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: log(average_RT) ~ 1 + typicality_condition_c + (1 | sub_num)
##    Data: avg_subj_RT
## 
## REML criterion at convergence: 151.3
## 
## Scaled residuals: 
##      Min       1Q   Median       3Q      Max 
## -1.94272 -0.53780  0.02645  0.58890  1.78915 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.05555  0.2357  
##  Residual             0.06904  0.2628  
## Number of obs: 219, groups:  sub_num, 119
## 
## Fixed effects:
##                          Estimate Std. Error         df t value Pr(>|t|)    
## (Intercept)              6.692444   0.028147 115.078188 237.765   <2e-16 ***
## typicality_condition_c  -0.007352   0.036226 106.523802  -0.203     0.84    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## typclty_cn_ -0.020

confint(m_4_1,method="Wald")

##                              2.5 %     97.5 %
## .sig01                          NA         NA
## .sigma                          NA         NA
## (Intercept)             6.63727671 6.74761208
## typicality_condition_c -0.07835359 0.06364991

There is no significant typicality effect.

4.3.1.1 RT for typical

m_4_1_1 <- lmer(log(average_RT) ~ 1 + typicality_condition_typ + (1|sub_num),data=avg_subj_RT)

summary(m_4_1_1)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: log(average_RT) ~ 1 + typicality_condition_typ + (1 | sub_num)
##    Data: avg_subj_RT
## 
## REML criterion at convergence: 151.3
## 
## Scaled residuals: 
##      Min       1Q   Median       3Q      Max 
## -1.94272 -0.53780  0.02645  0.58890  1.78915 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.05555  0.2357  
##  Residual             0.06904  0.2628  
## Number of obs: 219, groups:  sub_num, 119
## 
## Fixed effects:
##                            Estimate Std. Error         df t value Pr(>|t|)    
## (Intercept)                6.688768   0.033159 189.134101 201.720   <2e-16 ***
## typicality_condition_typ  -0.007352   0.036226 106.523802  -0.203     0.84    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## typclty_cn_ 0.529

confint(m_4_1_1,method="Wald")

##                                2.5 %     97.5 %
## .sig01                            NA         NA
## .sigma                            NA         NA
## (Intercept)               6.62377866 6.75375830
## typicality_condition_typ -0.07835359 0.06364991

RT for atypical

m_4_1_2 <- lmer(log(average_RT) ~ 1 + typicality_condition_atyp + (1|sub_num),data=avg_subj_RT)

summary(m_4_1_2)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: log(average_RT) ~ 1 + typicality_condition_atyp + (1 | sub_num)
##    Data: avg_subj_RT
## 
## REML criterion at convergence: 151.3
## 
## Scaled residuals: 
##      Min       1Q   Median       3Q      Max 
## -1.94272 -0.53780  0.02645  0.58890  1.78915 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.05555  0.2357  
##  Residual             0.06904  0.2628  
## Number of obs: 219, groups:  sub_num, 119
## 
## Fixed effects:
##                             Estimate Std. Error         df t value Pr(>|t|)    
## (Intercept)                 6.696120   0.033782 193.156112 198.217   <2e-16 ***
## typicality_condition_atyp  -0.007352   0.036226 106.523802  -0.203     0.84    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## typclty_cn_ -0.553

confint(m_4_1_2,method="Wald")

##                                 2.5 %     97.5 %
## .sig01                             NA         NA
## .sigma                             NA         NA
## (Intercept)                6.62990934 6.76233130
## typicality_condition_atyp -0.07835359 0.06364991

4.3.2 Trial level RT

trial_rt <- d_rt %>%
  filter(shift_type=="D-T")%>%
  filter(useable_trials=="1") %>%
  mutate(
    typicality_condition_c = case_when(
      condition == "atypical" ~ -0.5,
      condition == "typical" ~ 0.5,
      TRUE ~ NA_real_
    ),
    typicality_condition_typ = case_when(
      condition == "atypical" ~ -1,
      condition == "typical" ~ 0,
      TRUE ~ NA_real_
    ),
    typicality_condition_atyp = case_when(
      condition == "atypical" ~ 0,
      condition == "typical" ~ 1,
      TRUE ~ NA_real_
    ),
  )

m_4_2 <- lmer(log(rt) ~ 1 + typicality_condition_c + 
            (1+ typicality_condition_c||sub_num) +
            (1|category),
          data=trial_rt)

summary(m_4_2)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: log(rt) ~ 1 + typicality_condition_c + (1 + typicality_condition_c ||  
##     sub_num) + (1 | category)
##    Data: trial_rt
## 
## REML criterion at convergence: 4289.5
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -2.7478 -0.6470  0.1102  0.6435  2.3437 
## 
## Random effects:
##  Groups    Name                   Variance Std.Dev.
##  sub_num   (Intercept)            0.028568 0.1690  
##  sub_num.1 typicality_condition_c 0.000000 0.0000  
##  category  (Intercept)            0.003147 0.0561  
##  Residual                         0.623377 0.7895  
## Number of obs: 1783, groups:  sub_num, 119; category, 4
## 
## Fixed effects:
##                         Estimate Std. Error        df t value Pr(>|t|)    
## (Intercept)            6.424e+00  3.747e-02 4.515e+00 171.437 8.75e-10 ***
## typicality_condition_c 7.012e-03  3.783e-02 1.758e+03   0.185    0.853    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## typclty_cn_ 0.000 
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

confint(m_4_2,method="Wald")

##                              2.5 %     97.5 %
## .sig01                          NA         NA
## .sig02                          NA         NA
## .sig03                          NA         NA
## .sigma                          NA         NA
## (Intercept)             6.35065445 6.49754253
## typicality_condition_c -0.06713151 0.08115617

4.3 Aim 3 models with RT as the outcome

Does experience with exemplars predict reaction time?

parent_typicality_rt <- parent_typicality_z %>%
  #filter(exclude_participant==0) %>%
  #filter(useable_window==1) %>%
  distinct(sub_num,months,age,child_gender, trial_order,trial_number,category,target_image,target_typicality_z,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking,target_parent_typicality_z,distractor_parent_typicality_z,target_image,target_parent_typicality_rating,distractor_parent_typicality_rating)

trial_rt_parent <- left_join(trial_rt, parent_typicality_rt)  
  
trial_rt_parent <- trial_rt_parent%>%
  filter(shift_type == "D-T") %>%
  filter(useable_trials == "1")%>%
  mutate(
    age_mo_c = age_mo - mean(age_mo),
    age_c = age - mean(age)
  )

##subject details for aim 3 analysis with RT
rt_survey_subject_info <- parent_typicality_rt %>%
  ungroup()%>%
  summarize(
    N = length(unique(sub_num)),
    mean_age = mean(age),
    sd_age = sd(age)
  )

rt_survey_subject_info%>%
  knitr::kable()

N	mean_age	sd_age
133	479.8362	46.56385

#model
m_4_3 <- lmer(log(rt) ~ target_parent_typicality_z+age_c+(target_parent_typicality_z|sub_num) + (1|category), trial_rt_parent)

summary(m_4_3)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: 
## log(rt) ~ target_parent_typicality_z + age_c + (target_parent_typicality_z |  
##     sub_num) + (1 | category)
##    Data: trial_rt_parent
## 
## REML criterion at convergence: 3551.7
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -2.7734 -0.6431  0.1127  0.6559  2.3357 
## 
## Random effects:
##  Groups   Name                       Variance  Std.Dev. Corr 
##  sub_num  (Intercept)                2.922e-02 0.170937      
##           target_parent_typicality_z 3.640e-06 0.001908 -1.00
##  category (Intercept)                4.251e-03 0.065198      
##  Residual                            6.119e-01 0.782244      
## Number of obs: 1481, groups:  sub_num, 93; category, 4
## 
## Fixed effects:
##                              Estimate Std. Error         df t value Pr(>|t|)
## (Intercept)                 6.399e+00  4.284e-02  4.519e+00 149.362  1.6e-09
## target_parent_typicality_z -1.020e-03  2.079e-02  1.392e+03  -0.049    0.961
## age_c                      -9.308e-04  6.144e-04  8.656e+01  -1.515    0.133
##                               
## (Intercept)                ***
## target_parent_typicality_z    
## age_c                         
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) trg___
## trgt_prnt__  0.012       
## age_c        0.084 -0.008
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

confint(m_4_3,method="Wald")

##                                  2.5 %       97.5 %
## .sig01                              NA           NA
## .sig02                              NA           NA
## .sig03                              NA           NA
## .sig04                              NA           NA
## .sigma                              NA           NA
## (Intercept)                 6.31469153 6.4826212542
## target_parent_typicality_z -0.04176220 0.0397222921
## age_c                      -0.00213508 0.0002734553

Parent report of typicality does not predict RT.

4.4 Do effects interact with test session?

We will fit models analogous to those in 1.1 and 1.2 including an interaction with test session

4.4.1.1

#baseline-corrected target looking by session
avg_corrected_target_looking_by_typicality_session <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num, age,age_mo, child_gender, trial_order,trial_number,target_image,target_typicality_z,condition,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking,session_num) %>%
  group_by(sub_num, child_gender,age,age_mo,condition, session_num) %>%
  summarize(N=n(),
            average_corrected_target_looking=mean(corrected_target_looking,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking,na.rm=T)/sqrt(N),
            lower_ci=average_corrected_target_looking-ci,
            upper_ci=average_corrected_target_looking+ci)


avg_corrected_target_looking_by_typicality_session <- avg_corrected_target_looking_by_typicality_session %>%
  mutate(
    typicality_condition_c = case_when(
      condition == "atypical" ~ -0.5,
      condition == "typical" ~ 0.5,
      TRUE ~ NA_real_
    ),
    typicality_condition_typ = case_when(
      condition == "atypical" ~ -1,
      condition == "typical" ~ 0,
      TRUE ~ NA_real_
    ),
    typicality_condition_atyp = case_when(
      condition == "atypical" ~ 0,
      condition == "typical" ~ 1,
      TRUE ~ NA_real_
    ),    
    session_num_c = case_when(
      session_num == "1" ~ -0.5,
      session_num == "2" ~ 0.5,
      TRUE ~ NA_real_
    )
  )

m_4_4_1_1 <- lmer(average_corrected_target_looking ~ 1 + typicality_condition_c*session_num_c + (1|sub_num),data=avg_corrected_target_looking_by_typicality_session)

summary(m_4_4_1_1)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: average_corrected_target_looking ~ 1 + typicality_condition_c *  
##     session_num_c + (1 | sub_num)
##    Data: avg_corrected_target_looking_by_typicality_session
## 
## REML criterion at convergence: -403.9
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.6162 -0.6597 -0.0075  0.6086  3.8622 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.002852 0.0534  
##  Residual             0.014118 0.1188  
## Number of obs: 331, groups:  sub_num, 85
## 
## Fixed effects:
##                                       Estimate Std. Error        df t value
## (Intercept)                            0.08150    0.02528 171.77267   3.224
## typicality_condition_c                 0.03719    0.04253 243.12809   0.875
## session_num_c                         -0.02750    0.05056 171.77267  -0.544
## typicality_condition_c:session_num_c  -0.02803    0.08505 243.12809  -0.330
##                                      Pr(>|t|)   
## (Intercept)                           0.00151 **
## typicality_condition_c                0.38268   
## session_num_c                         0.58724   
## typicality_condition_c:session_num_c  0.74198   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) typc__ sssn__
## typclty_cn_  0.000              
## sessin_nm_c -0.938  0.000       
## typclt__:__  0.000 -0.952  0.000

confint(m_4_4_1_1,method="Wald")

##                                            2.5 %     97.5 %
## .sig01                                        NA         NA
## .sigma                                        NA         NA
## (Intercept)                           0.03195216 0.13105448
## typicality_condition_c               -0.04615820 0.12054049
## session_num_c                        -0.12660176 0.07160286
## typicality_condition_c:session_num_c -0.19473211 0.13866527

4.4.1.1

No significant interactions with test session

4.4.1.1.3

m_4_4_1_1_3_typ <- lmer(average_corrected_target_looking ~ 1 + typicality_condition_typ*session_num_c + (1|sub_num),data=avg_corrected_target_looking_by_typicality_session)

summary(m_4_4_1_1_3_typ)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: average_corrected_target_looking ~ 1 + typicality_condition_typ *  
##     session_num_c + (1 | sub_num)
##    Data: avg_corrected_target_looking_by_typicality_session
## 
## REML criterion at convergence: -403.9
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.6162 -0.6597 -0.0075  0.6086  3.8622 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.002852 0.0534  
##  Residual             0.014118 0.1188  
## Number of obs: 331, groups:  sub_num, 85
## 
## Fixed effects:
##                                         Estimate Std. Error        df t value
## (Intercept)                              0.10010    0.03304 297.75336   3.030
## typicality_condition_typ                 0.03719    0.04253 243.12809   0.875
## session_num_c                           -0.04152    0.06607 297.75336  -0.628
## typicality_condition_typ:session_num_c  -0.02803    0.08505 243.12809  -0.330
##                                        Pr(>|t|)   
## (Intercept)                             0.00266 **
## typicality_condition_typ                0.38268   
## session_num_c                           0.53025   
## typicality_condition_typ:session_num_c  0.74198   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) typc__ sssn__
## typclty_cn_  0.644              
## sessin_nm_c -0.944 -0.612       
## typclt__:__ -0.612 -0.952  0.644

confint(m_4_4_1_1_3_typ,method="Wald")

##                                              2.5 %     97.5 %
## .sig01                                          NA         NA
## .sigma                                          NA         NA
## (Intercept)                             0.03535024 0.16484755
## typicality_condition_typ               -0.04615820 0.12054049
## session_num_c                          -0.17101346 0.08798114
## typicality_condition_typ:session_num_c -0.19473211 0.13866527

Infants successfully recognize words in the typical condition.

m_4_4_1_1_3_atyp <- lmer(average_corrected_target_looking ~ 1 + typicality_condition_atyp*session_num_c + (1|sub_num),data=avg_corrected_target_looking_by_typicality_session)

summary(m_4_4_1_1_3_atyp)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: average_corrected_target_looking ~ 1 + typicality_condition_atyp *  
##     session_num_c + (1 | sub_num)
##    Data: avg_corrected_target_looking_by_typicality_session
## 
## REML criterion at convergence: -403.9
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.6162 -0.6597 -0.0075  0.6086  3.8622 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.002852 0.0534  
##  Residual             0.014118 0.1188  
## Number of obs: 331, groups:  sub_num, 85
## 
## Fixed effects:
##                                          Estimate Std. Error        df t value
## (Intercept)                               0.06291    0.03303 297.74062   1.904
## typicality_condition_atyp                 0.03719    0.04253 243.12809   0.875
## session_num_c                            -0.01348    0.06607 297.74062  -0.204
## typicality_condition_atyp:session_num_c  -0.02803    0.08505 243.12809  -0.330
##                                         Pr(>|t|)  
## (Intercept)                               0.0578 .
## typicality_condition_atyp                 0.3827  
## session_num_c                             0.8384  
## typicality_condition_atyp:session_num_c   0.7420  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) typc__ sssn__
## typclty_cn_ -0.644              
## sessin_nm_c -0.944  0.613       
## typclt__:__  0.613 -0.952 -0.644

confint(m_4_4_1_1_3_atyp,method="Wald")

##                                                2.5 %    97.5 %
## .sig01                                            NA        NA
## .sigma                                            NA        NA
## (Intercept)                             -0.001836388 0.1276519
## typicality_condition_atyp               -0.046158198 0.1205405
## session_num_c                           -0.142971007 0.1160055
## typicality_condition_atyp:session_num_c -0.194732113 0.1386653

For an average session, infants recognition in the atypical condition is only marginally significantly above chance.

4.4.1.2

#baseline-corrected target looking
## trial-level
trial_corrected_accuracy_session <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num,session, age,age_mo, child_gender, trial_order,trial_number,category,target_image,target_typicality_z, condition,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking, session_num)


trial_corrected_accuracy_session <- trial_corrected_accuracy_session %>%
  mutate(
    typicality_condition_c = case_when(
      condition == "atypical" ~ -0.5,
      condition == "typical" ~ 0.5,
      TRUE ~ NA_real_
    ),
    typicality_condition_typ = case_when(
      condition == "atypical" ~ -1,
      condition == "typical" ~ 0,
      TRUE ~ NA_real_
    ),
    typicality_condition_atyp = case_when(
      condition == "atypical" ~ 0,
      condition == "typical" ~ 1,
      TRUE ~ NA_real_
    ),
    session_num_c = case_when(
      session_num == "1" ~ -0.5,
      session_num == "2" ~ 0.5,
      TRUE ~ NA_real_
    )
  )


m_4_4_1_2 <- lmer(corrected_target_looking ~ 1 + typicality_condition_c*session_num_c + 
            (1 + typicality_condition_c||sub_num) +
            (1|category),
          data=trial_corrected_accuracy_session)
summary(m_4_4_1_2)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: 
## corrected_target_looking ~ 1 + typicality_condition_c * session_num_c +  
##     (1 + typicality_condition_c || sub_num) + (1 | category)
##    Data: trial_corrected_accuracy_session
## 
## REML criterion at convergence: 2301.5
## 
## Scaled residuals: 
##      Min       1Q   Median       3Q      Max 
## -3.14509 -0.63228 -0.02488  0.66678  2.76849 
## 
## Random effects:
##  Groups    Name                   Variance  Std.Dev.
##  sub_num   (Intercept)            0.0025671 0.05067 
##  sub_num.1 typicality_condition_c 0.0000000 0.00000 
##  category  (Intercept)            0.0001653 0.01286 
##  Residual                         0.1192192 0.34528 
## Number of obs: 3138, groups:  sub_num, 85; category, 4
## 
## Fixed effects:
##                                        Estimate Std. Error         df t value
## (Intercept)                             0.08227    0.02302   72.73235   3.573
## typicality_condition_c                  0.03337    0.03579 3049.70767   0.932
## session_num_c                          -0.02596    0.04422  137.21796  -0.587
## typicality_condition_c:session_num_c   -0.03567    0.07159 3049.70767  -0.498
##                                      Pr(>|t|)    
## (Intercept)                           0.00063 ***
## typicality_condition_c                0.35118    
## session_num_c                         0.55813    
## typicality_condition_c:session_num_c  0.61836    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) typc__ sssn__
## typclty_cn_  0.000              
## sessin_nm_c -0.890  0.000       
## typclt__:__  0.000 -0.939  0.000
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

Exploratory analyses:

Ex1. typicality as a likert scale

We intentionally selected items at the end of a typicality scale from a set of adult-normed images. We categorized each image as typical or atypical based on adults ratings, however, we also have the origninal normed values. Does typicality influence accuracy when typicality is a continuum rather than a dichotomous variable.

By Continuous typicality

Here, we summarize each participants’ average accuracy during the critical window and average baseline-corrected proportion target looking.

# critical window only
avg_critical_window_accuracy_by_typicality_continuous <- d %>%
  filter(exclude_participant_critical==0) %>%
  filter(useable_critical_window==1) %>%
  filter(corrected_time_centered>=300&corrected_time_centered<=2800) %>%
  group_by(sub_num, age,age_mo, child_gender, trial_order,trial_number,target_image,target_typicality_z) %>%
  summarize(mean_accuracy=mean(accuracy_transformed,na.rm=TRUE)) %>%
  ungroup() %>%
  group_by(sub_num, child_gender,target_typicality_z) %>%
  summarize(N=n(),
            mean_age = mean(age),
            mean_age_mo = mean(age_mo),
            accuracy=mean(mean_accuracy,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(mean_accuracy,na.rm=T)/sqrt(N),
            lower_ci=accuracy-ci,
            upper_ci=accuracy+ci)

#baseline-corrected target looking
avg_corrected_target_looking_by_typicality_continuous <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num, age,age_mo, child_gender, trial_order,trial_number,target_image,target_typicality_z,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking) %>%
  group_by(sub_num, child_gender,age,age_mo,target_typicality_z) %>%
  summarize(N=n(),
            average_corrected_target_looking=mean(corrected_target_looking,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking,na.rm=T)/sqrt(N),
            lower_ci=average_corrected_target_looking-ci,
            upper_ci=average_corrected_target_looking+ci)

#avg typicality baseline-corrected target looking
avg_corrected_target_looking_category_continuous <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num, months,age_mo,child_gender, trial_order,trial_number,category,target_image,target_typicality_z,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking) %>%
  group_by(sub_num, age_mo,child_gender, trial_order,target_typicality_z,category) %>%
  summarize(N=n(),
            average_corrected_target_looking=mean(corrected_target_looking,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking,na.rm=T)/sqrt(N),
            lower_ci=average_corrected_target_looking-ci,
            upper_ci=average_corrected_target_looking+ci)

#avg typicality baseline-corrected target looking
avg_corrected_target_looking_typicality_continuous <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num, months,age_mo,child_gender, trial_order,trial_number,target_image,target_typicality_z,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking) %>%
  group_by(sub_num, age_mo,child_gender, trial_order,target_typicality_z) %>%
  summarize(N=n(),
            average_corrected_target_looking=mean(corrected_target_looking,na.rm=TRUE),
            ci=qt(0.975, N-1)*sd(corrected_target_looking,na.rm=T)/sqrt(N),
            se=sd(corrected_target_looking,na.rm=T)/sqrt(N),
            lower_ci=average_corrected_target_looking-ci,
            upper_ci=average_corrected_target_looking+ci,
            lower_se=average_corrected_target_looking-se,
            upper_se=average_corrected_target_looking+se)

#Subject level

m_Ex1_1 <- lmer(average_corrected_target_looking ~ 1 + target_typicality_z + (1|sub_num),data=avg_corrected_target_looking_by_typicality_continuous)

summary(m_Ex1_1)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: average_corrected_target_looking ~ 1 + target_typicality_z +  
##     (1 | sub_num)
##    Data: avg_corrected_target_looking_by_typicality_continuous
## 
## REML criterion at convergence: 2293.8
## 
## Scaled residuals: 
##      Min       1Q   Median       3Q      Max 
## -3.15466 -0.63486 -0.01697  0.67445  2.77538 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  sub_num  (Intercept) 0.002532 0.05032 
##  Residual             0.119198 0.34525 
## Number of obs: 3138, groups:  sub_num, 85
## 
## Fixed effects:
##                      Estimate Std. Error        df t value Pr(>|t|)    
## (Intercept)         7.029e-02  8.270e-03 8.357e+01   8.499 6.27e-13 ***
## target_typicality_z 1.371e-02  6.271e-03 3.067e+03   2.185   0.0289 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## trgt_typcl_ 0.004

confint(m_Ex1_1,method="Wald")

##                           2.5 %     97.5 %
## .sig01                       NA         NA
## .sigma                       NA         NA
## (Intercept)         0.054078391 0.08649702
## target_typicality_z 0.001413998 0.02599765

When we use the full scale, rather than a dichotomous scale, typicality is a significant predictor of word recognition (p=0.029)

#Trial Level

m_Ex1_2 <- lmer(corrected_target_looking ~ 1 + target_typicality_z + 
            (1 + target_typicality_z||sub_num) +
            (1|category),
          data=trial_corrected_accuracy)
summary(m_Ex1_2)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: 
## corrected_target_looking ~ 1 + target_typicality_z + (1 + target_typicality_z ||  
##     sub_num) + (1 | category)
##    Data: trial_corrected_accuracy
## 
## REML criterion at convergence: 2292.4
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.1663 -0.6275 -0.0227  0.6732  2.7629 
## 
## Random effects:
##  Groups    Name                Variance  Std.Dev.
##  sub_num   (Intercept)         0.0025456 0.05045 
##  sub_num.1 target_typicality_z 0.0000000 0.00000 
##  category  (Intercept)         0.0001945 0.01395 
##  Residual                      0.1190434 0.34503 
## Number of obs: 3138, groups:  sub_num, 85; category, 4
## 
## Fixed effects:
##                      Estimate Std. Error        df t value Pr(>|t|)    
## (Intercept)         7.034e-02  1.082e-02 5.347e+00   6.499 0.000996 ***
## target_typicality_z 1.414e-02  6.304e-03 2.872e+03   2.244 0.024923 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr)
## trgt_typcl_ 0.003 
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

confint(m_Ex1_2,method="Wald")

##                           2.5 %     97.5 %
## .sig01                       NA         NA
## .sig02                       NA         NA
## .sig03                       NA         NA
## .sigma                       NA         NA
## (Intercept)         0.049124287 0.09155134
## target_typicality_z 0.001789097 0.02649962

##Ex2. Does typicality interact with category? Infants’ processing for some categories may be better than other categories. Are infants better at processing typical category members for categories that are more accurate?

trial_corrected_accuracy <- trial_corrected_accuracy %>%
  mutate(
    category_c1 = case_when(
      category == "bird" ~ 1,
      category == "cat" ~ 0,
      category == "dog" ~ 0,
      category == "fish" ~ 0,
      TRUE ~ NA_real_
    ),
    category_c2 = case_when(
      category == "bird" ~ 0,
      category == "cat" ~ 1,
      category == "dog" ~ 0,
      category == "fish" ~ 0,
      TRUE ~ NA_real_
    ),
    category_c3 = case_when(
      category == "bird" ~ 0,
      category == "cat" ~ 0,
      category == "dog" ~ 0,
      category == "fish" ~ 1,
      TRUE ~ NA_real_
    )
  )

m_Ex2_1 <- lme4::lmer(corrected_target_looking ~ typicality_condition_c*(category_c1+category_c2+category_c3)+
            (typicality_condition_c*(category_c1+category_c2+category_c3)||sub_num),control=lmerControl(optCtrl=list(maxfun=50000),optimizer = "bobyqa"),
          data=trial_corrected_accuracy)

m_Ex2_C <- lme4::lmer(corrected_target_looking ~ typicality_condition_c+category_c1+category_c2+category_c3+
            (typicality_condition_c+category_c1+category_c2+category_c3|sub_num),control=lmerControl(optCtrl=list(maxfun=50000),optimizer = "bobyqa"),
          data=trial_corrected_accuracy)

#modelCompare(m_Ex2_C,m_Ex2_1)

summary(m_Ex2_1)

## Linear mixed model fit by REML ['lmerMod']
## Formula: corrected_target_looking ~ typicality_condition_c * (category_c1 +  
##     category_c2 + category_c3) + ((1 | sub_num) + (0 + typicality_condition_c |  
##     sub_num) + (0 + category_c1 | sub_num) + (0 + category_c2 |  
##     sub_num) + (0 + category_c3 | sub_num) + (0 + typicality_condition_c:category_c1 |  
##     sub_num) + (0 + typicality_condition_c:category_c2 | sub_num) +  
##     (0 + typicality_condition_c:category_c3 | sub_num))
##    Data: trial_corrected_accuracy
## Control: lmerControl(optCtrl = list(maxfun = 50000), optimizer = "bobyqa")
## 
## REML criterion at convergence: 2312.8
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.2277 -0.6266 -0.0182  0.6588  2.8173 
## 
## Random effects:
##  Groups    Name                               Variance  Std.Dev.
##  sub_num   (Intercept)                        0.0020271 0.04502 
##  sub_num.1 typicality_condition_c             0.0000000 0.00000 
##  sub_num.2 category_c1                        0.0008921 0.02987 
##  sub_num.3 category_c2                        0.0009847 0.03138 
##  sub_num.4 category_c3                        0.0037239 0.06102 
##  sub_num.5 typicality_condition_c:category_c1 0.0000000 0.00000 
##  sub_num.6 typicality_condition_c:category_c2 0.0029365 0.05419 
##  sub_num.7 typicality_condition_c:category_c3 0.0119026 0.10910 
##  Residual                                     0.1171444 0.34226 
## Number of obs: 3138, groups:  sub_num, 85
## 
## Fixed effects:
##                                    Estimate Std. Error t value
## (Intercept)                         0.04810    0.01316   3.654
## typicality_condition_c              0.05311    0.02441   2.176
## category_c1                         0.03428    0.01769   1.938
## category_c2                         0.03881    0.01762   2.202
## category_c3                         0.01614    0.01849   0.873
## typicality_condition_c:category_c1 -0.07938    0.03477  -2.283
## typicality_condition_c:category_c2 -0.01219    0.03509  -0.348
## typicality_condition_c:category_c3 -0.05654    0.03649  -1.550
## 
## Correlation of Fixed Effects:
##             (Intr) typc__ ctgr_1 ctgr_2 ctgr_3 t__:_1 t__:_2
## typclty_cn_  0.018                                          
## category_c1 -0.640 -0.014                                   
## category_c2 -0.642 -0.014  0.478                            
## category_c3 -0.612 -0.013  0.455  0.457                     
## typclt__:_1 -0.013 -0.702 -0.008  0.009  0.009              
## typclt__:_2 -0.013 -0.696  0.010  0.008  0.009  0.489       
## typclt__:_3 -0.012 -0.669  0.009  0.009  0.012  0.470  0.466
## optimizer (bobyqa) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')

confint(m_Ex2_1,method="Wald")

##                                            2.5 %      97.5 %
## .sig01                                        NA          NA
## .sig02                                        NA          NA
## .sig03                                        NA          NA
## .sig04                                        NA          NA
## .sig05                                        NA          NA
## .sig06                                        NA          NA
## .sig07                                        NA          NA
## .sig08                                        NA          NA
## .sigma                                        NA          NA
## (Intercept)                         0.0223006277  0.07389738
## typicality_condition_c              0.0052673224  0.10095736
## category_c1                        -0.0003881031  0.06893827
## category_c2                         0.0042713873  0.07335071
## category_c3                        -0.0200975714  0.05237435
## typicality_condition_c:category_c1 -0.1475284257 -0.01123941
## typicality_condition_c:category_c2 -0.0809618648  0.05657510
## typicality_condition_c:category_c3 -0.1280508943  0.01497452

Ex3. Does a typicality exert more of an effect on processing when the categories tested are more similar?

There are some categories that are more visually/categorically similar than other categories (i.e., cats and dogs vs. dogs and fish). Is there an effect of typicality when there is more category similarity?

trial_corrected_accuracy_category <- d %>%
  filter(exclude_participant==0) %>%
  filter(useable_window==1) %>%
  distinct(sub_num,session, age,age_mo, child_gender, trial_order,trial_number,category,left_image,right_image,target_image,target_typicality_z, condition,mean_target_looking_critical,mean_target_looking_baseline,corrected_target_looking)%>%
  group_by(
    category,
    target_image
  )